/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006, 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2009, 2010, 2011, 2012, 2013, 2014, 2016 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
/* FIXME:
- - Pearson's R (but not Spearman!) is off a little.
- - T values for Spearman's R and Pearson's R are wrong.
- - How to calculate significance of symmetric and directional measures?
- - Asymmetric ASEs and T values for lambda are wrong.
- - ASE of Goodman and Kruskal's tau is not calculated.
- - ASE of symmetric somers' d is wrong.
- - Approx. T of uncertainty coefficient is wrong.
+ - How to calculate significance of some symmetric and directional measures?
+ - How to calculate ASE for symmetric Somers ' d?
+ - How to calculate ASE for Goodman and Kruskal's tau?
+ - How to calculate approx. T of symmetric uncertainty coefficient?
*/
#include <config.h>
#include <ctype.h>
+#include <float.h>
#include <gsl/gsl_cdf.h>
#include <stdlib.h>
#include <stdio.h>
#include "data/value-labels.h"
#include "data/variable.h"
#include "language/command.h"
+#include "language/stats/freq.h"
#include "language/dictionary/split-file.h"
#include "language/lexer/lexer.h"
#include "language/lexer/variable-parser.h"
#include "libpspp/pool.h"
#include "libpspp/str.h"
#include "output/tab.h"
+#include "output/chart-item.h"
+#include "output/charts/barchart.h"
#include "gl/minmax.h"
#include "gl/xalloc.h"
*^tables=custom;
+variables=custom;
missing=miss:!table/include/report;
+ count=roundwhat:asis/case/!cell,
+ roundhow:!round/truncate;
+write[wr_]=none,cells,all;
+format=val:!avalue/dvalue,
indx:!noindex/index,
tabl:!tables/notables,
box:!box/nobox,
pivot:!pivot/nopivot;
+ +barchart=;
+cells[cl_]=count,expected,row,column,total,residual,sresidual,
asresidual,all,none;
+statistics[st_]=chisq,phi,cc,lambda,uc,none,btau,ctau,risk,gamma,d,
/* Number of directional statistics. */
#define N_DIRECTIONAL 13
-/* A single table entry for general mode. */
-struct table_entry
- {
- struct hmap_node node; /* Entry in hash table. */
- double freq; /* Frequency count. */
- union value values[1]; /* Values. */
- };
-
-static size_t
-table_entry_size (size_t n_values)
-{
- return (offsetof (struct table_entry, values)
- + n_values * sizeof (union value));
-}
-/* Indexes into the 'vars' member of struct pivot_table and
+/* Indexes into the 'vars' member of struct crosstabulation and
struct crosstab member. */
enum
{
/* Higher indexes cause multiple tables to be output. */
};
+struct xtab_var
+ {
+ const struct variable *var;
+ union value *values;
+ size_t n_values;
+ };
+
/* A crosstabulation of 2 or more variables. */
-struct pivot_table
+struct crosstabulation
{
+ struct crosstabs_proc *proc;
struct fmt_spec weight_format; /* Format for weight variable. */
double missing; /* Weight of missing cases. */
/* Variables (2 or more). */
int n_vars;
- const struct variable **vars;
+ struct xtab_var *vars;
/* Constants (0 or more). */
int n_consts;
- const struct variable **const_vars;
- union value *const_values;
+ struct xtab_var *const_vars;
/* Data. */
struct hmap data;
- struct table_entry **entries;
+ struct freq **entries;
size_t n_entries;
- /* Column values, number of columns. */
- union value *cols;
- int n_cols;
-
- /* Row values, number of rows. */
- union value *rows;
- int n_rows;
-
/* Number of statistically interesting columns/rows
(columns/rows with data in them). */
int ns_cols, ns_rows;
/* Integer mode variable info. */
struct var_range
{
+ struct hmap_node hmap_node; /* In struct crosstabs_proc var_ranges map. */
+ const struct variable *var; /* The variable. */
int min; /* Minimum value. */
int max; /* Maximum value + 1. */
int count; /* max - min. */
};
-static inline struct var_range *
-get_var_range (const struct variable *v)
-{
- return var_get_aux (v);
-}
-
struct crosstabs_proc
{
const struct dictionary *dict;
enum { INTEGER, GENERAL } mode;
enum mv_class exclude;
bool pivot;
+ bool barchart;
bool bad_warn;
struct fmt_spec weight_format;
/* Variables specifies on VARIABLES. */
const struct variable **variables;
size_t n_variables;
+ struct hmap var_ranges;
/* TABLES. */
- struct pivot_table *pivots;
+ struct crosstabulation *pivots;
int n_pivots;
/* CELLS. */
unsigned int cells; /* Bit k is 1 if cell k is requested. */
int a_cells[CRS_CL_count]; /* 0...n_cells-1 are the requested cells. */
+ /* Rounding of cells. */
+ bool round_case_weights; /* Round case weights? */
+ bool round_cells; /* If !round_case_weights, round cells? */
+ bool round_down; /* Round down? (otherwise to nearest) */
+
/* STATISTICS. */
unsigned int statistics; /* Bit k is 1 if statistic k is requested. */
bool descending; /* True if descending sort order is requested. */
};
-static bool should_tabulate_case (const struct pivot_table *,
+const struct var_range *get_var_range (const struct crosstabs_proc *,
+ const struct variable *);
+
+static bool should_tabulate_case (const struct crosstabulation *,
const struct ccase *, enum mv_class exclude);
-static void tabulate_general_case (struct pivot_table *, const struct ccase *,
+static void tabulate_general_case (struct crosstabulation *, const struct ccase *,
double weight);
-static void tabulate_integer_case (struct pivot_table *, const struct ccase *,
+static void tabulate_integer_case (struct crosstabulation *, const struct ccase *,
double weight);
static void postcalc (struct crosstabs_proc *);
-static void submit (struct pivot_table *, struct tab_table *);
+static void submit (struct crosstabulation *, struct tab_table *);
+
+static double
+round_weight (const struct crosstabs_proc *proc, double weight)
+{
+ return proc->round_down ? floor (weight) : floor (weight + 0.5);
+}
/* Parses and executes the CROSSTABS procedure. */
int
cmd_crosstabs (struct lexer *lexer, struct dataset *ds)
{
- const struct variable *wv = dict_get_weight (dataset_dict (ds));
+ struct var_range *range, *next_range;
struct crosstabs_proc proc;
struct casegrouper *grouper;
struct casereader *input, *group;
struct cmd_crosstabs cmd;
- struct pivot_table *pt;
+ struct crosstabulation *xt;
int result;
bool ok;
int i;
proc.bad_warn = true;
proc.variables = NULL;
proc.n_variables = 0;
+ hmap_init (&proc.var_ranges);
proc.pivots = NULL;
proc.n_pivots = 0;
proc.descending = false;
- proc.weight_format = wv ? *var_get_print_format (wv) : F_8_0;
+ proc.weight_format = *dict_get_weight_format (dataset_dict (ds));
if (!parse_crosstabs (lexer, ds, &cmd, &proc))
{
}
proc.mode = proc.n_variables ? INTEGER : GENERAL;
-
+ proc.barchart = cmd.sbc_barchart > 0;
proc.descending = cmd.val == CRS_DVALUE;
+ proc.round_case_weights = cmd.sbc_count && cmd.roundwhat == CRS_CASE;
+ proc.round_cells = cmd.sbc_count && cmd.roundwhat == CRS_CELL;
+ proc.round_down = cmd.roundhow == CRS_TRUNCATE;
+
/* CELLS. */
if (!cmd.sbc_cells)
proc.cells = 1u << CRS_CL_COUNT;
: MV_NEVER);
if (proc.mode == GENERAL && proc.exclude == MV_NEVER)
{
- msg (SE, _("Missing mode REPORT not allowed in general mode. "
- "Assuming MISSING=TABLE."));
+ msg (SE, _("Missing mode %s not allowed in general mode. "
+ "Assuming %s."), "REPORT", "MISSING=TABLE");
proc.exclude = MV_ANY;
}
}
/* Initialize hash tables. */
- for (pt = &proc.pivots[0]; pt < &proc.pivots[proc.n_pivots]; pt++)
- hmap_init (&pt->data);
+ for (xt = &proc.pivots[0]; xt < &proc.pivots[proc.n_pivots]; xt++)
+ hmap_init (&xt->data);
/* Tabulate. */
for (; (c = casereader_read (group)) != NULL; case_unref (c))
- for (pt = &proc.pivots[0]; pt < &proc.pivots[proc.n_pivots]; pt++)
+ for (xt = &proc.pivots[0]; xt < &proc.pivots[proc.n_pivots]; xt++)
{
double weight = dict_get_case_weight (dataset_dict (ds), c,
&proc.bad_warn);
- if (should_tabulate_case (pt, c, proc.exclude))
+ if (cmd.roundwhat == CRS_CASE)
+ {
+ weight = round_weight (&proc, weight);
+ if (weight == 0.)
+ continue;
+ }
+ if (should_tabulate_case (xt, c, proc.exclude))
{
if (proc.mode == GENERAL)
- tabulate_general_case (pt, c, weight);
+ tabulate_general_case (xt, c, weight);
else
- tabulate_integer_case (pt, c, weight);
+ tabulate_integer_case (xt, c, weight);
}
else
- pt->missing += weight;
+ xt->missing += weight;
}
casereader_destroy (group);
exit:
free (proc.variables);
- for (pt = &proc.pivots[0]; pt < &proc.pivots[proc.n_pivots]; pt++)
+ HMAP_FOR_EACH_SAFE (range, next_range, struct var_range, hmap_node,
+ &proc.var_ranges)
+ {
+ hmap_delete (&proc.var_ranges, &range->hmap_node);
+ free (range);
+ }
+ for (xt = &proc.pivots[0]; xt < &proc.pivots[proc.n_pivots]; xt++)
{
- free (pt->vars);
- free (pt->const_vars);
+ free (xt->vars);
/* We must not call value_destroy on const_values because
it is a wild pointer; it never pointed to anything owned
- by the pivot_table.
+ by the crosstabulation.
The rest of the data was allocated and destroyed at a
lower level already. */
if (!lex_match (lexer, T_BY))
{
if (n_by < 2)
- {
- lex_force_match (lexer, T_BY);
- goto done;
- }
+ goto done;
else
break;
}
proc->n_pivots + nx, sizeof *proc->pivots);
for (i = 0; i < nx; i++)
{
- struct pivot_table *pt = &proc->pivots[proc->n_pivots++];
+ struct crosstabulation *xt = &proc->pivots[proc->n_pivots++];
int j;
- pt->weight_format = proc->weight_format;
- pt->missing = 0.;
- pt->n_vars = n_by;
- pt->vars = xmalloc (n_by * sizeof *pt->vars);
- pt->n_consts = 0;
- pt->const_vars = NULL;
- pt->const_values = NULL;
+ xt->proc = proc;
+ xt->weight_format = proc->weight_format;
+ xt->missing = 0.;
+ xt->n_vars = n_by;
+ xt->vars = xcalloc (n_by, sizeof *xt->vars);
+ xt->n_consts = 0;
+ xt->const_vars = NULL;
for (j = 0; j < n_by; j++)
- pt->vars[j] = by[j][by_iter[j]];
+ xt->vars[j].var = by[j][by_iter[j]];
for (j = n_by - 1; j >= 0; j--)
{
struct crosstabs_proc *proc = proc_;
if (proc->n_pivots)
{
- msg (SE, _("VARIABLES must be specified before TABLES."));
+ msg (SE, _("%s must be specified before %s."), "VARIABLES", "TABLES");
return 0;
}
for (i = orig_nv; i < proc->n_variables; i++)
{
+ const struct variable *var = proc->variables[i];
struct var_range *vr = xmalloc (sizeof *vr);
+
+ vr->var = var;
vr->min = min;
- vr->max = max + 1.;
+ vr->max = max;
vr->count = max - min + 1;
- var_attach_aux (proc->variables[i], vr, var_dtor_free);
+ hmap_insert (&proc->var_ranges, &vr->hmap_node,
+ hash_pointer (var, 0));
}
if (lex_token (lexer) == T_SLASH)
\f
/* Data file processing. */
+const struct var_range *
+get_var_range (const struct crosstabs_proc *proc, const struct variable *var)
+{
+ if (!hmap_is_empty (&proc->var_ranges))
+ {
+ const struct var_range *range;
+
+ HMAP_FOR_EACH_IN_BUCKET (range, struct var_range, hmap_node,
+ hash_pointer (var, 0), &proc->var_ranges)
+ if (range->var == var)
+ return range;
+ }
+
+ return NULL;
+}
+
static bool
-should_tabulate_case (const struct pivot_table *pt, const struct ccase *c,
+should_tabulate_case (const struct crosstabulation *xt, const struct ccase *c,
enum mv_class exclude)
{
int j;
- for (j = 0; j < pt->n_vars; j++)
+ for (j = 0; j < xt->n_vars; j++)
{
- const struct variable *var = pt->vars[j];
- struct var_range *range = get_var_range (var);
+ const struct variable *var = xt->vars[j].var;
+ const struct var_range *range = get_var_range (xt->proc, var);
if (var_is_value_missing (var, case_data (c, var), exclude))
return false;
if (range != NULL)
{
double num = case_num (c, var);
- if (num < range->min || num > range->max)
+ if (num < range->min || num >= range->max + 1.)
return false;
}
}
}
static void
-tabulate_integer_case (struct pivot_table *pt, const struct ccase *c,
+tabulate_integer_case (struct crosstabulation *xt, const struct ccase *c,
double weight)
{
- struct table_entry *te;
+ struct freq *te;
size_t hash;
int j;
hash = 0;
- for (j = 0; j < pt->n_vars; j++)
+ for (j = 0; j < xt->n_vars; j++)
{
/* Throw away fractional parts of values. */
- hash = hash_int (case_num (c, pt->vars[j]), hash);
+ hash = hash_int (case_num (c, xt->vars[j].var), hash);
}
- HMAP_FOR_EACH_WITH_HASH (te, struct table_entry, node, hash, &pt->data)
+ HMAP_FOR_EACH_WITH_HASH (te, struct freq, node, hash, &xt->data)
{
- for (j = 0; j < pt->n_vars; j++)
- if ((int) case_num (c, pt->vars[j]) != (int) te->values[j].f)
+ for (j = 0; j < xt->n_vars; j++)
+ if ((int) case_num (c, xt->vars[j].var) != (int) te->values[j].f)
goto no_match;
/* Found an existing entry. */
- te->freq += weight;
+ te->count += weight;
return;
no_match: ;
}
/* No existing entry. Create a new one. */
- te = xmalloc (table_entry_size (pt->n_vars));
- te->freq = weight;
- for (j = 0; j < pt->n_vars; j++)
- te->values[j].f = (int) case_num (c, pt->vars[j]);
- hmap_insert (&pt->data, &te->node, hash);
+ te = xmalloc (table_entry_size (xt->n_vars));
+ te->count = weight;
+ for (j = 0; j < xt->n_vars; j++)
+ te->values[j].f = (int) case_num (c, xt->vars[j].var);
+ hmap_insert (&xt->data, &te->node, hash);
}
static void
-tabulate_general_case (struct pivot_table *pt, const struct ccase *c,
+tabulate_general_case (struct crosstabulation *xt, const struct ccase *c,
double weight)
{
- struct table_entry *te;
+ struct freq *te;
size_t hash;
int j;
hash = 0;
- for (j = 0; j < pt->n_vars; j++)
+ for (j = 0; j < xt->n_vars; j++)
{
- const struct variable *var = pt->vars[j];
+ const struct variable *var = xt->vars[j].var;
hash = value_hash (case_data (c, var), var_get_width (var), hash);
}
- HMAP_FOR_EACH_WITH_HASH (te, struct table_entry, node, hash, &pt->data)
+ HMAP_FOR_EACH_WITH_HASH (te, struct freq, node, hash, &xt->data)
{
- for (j = 0; j < pt->n_vars; j++)
+ for (j = 0; j < xt->n_vars; j++)
{
- const struct variable *var = pt->vars[j];
+ const struct variable *var = xt->vars[j].var;
if (!value_equal (case_data (c, var), &te->values[j],
var_get_width (var)))
goto no_match;
}
/* Found an existing entry. */
- te->freq += weight;
+ te->count += weight;
return;
no_match: ;
}
/* No existing entry. Create a new one. */
- te = xmalloc (table_entry_size (pt->n_vars));
- te->freq = weight;
- for (j = 0; j < pt->n_vars; j++)
+ te = xmalloc (table_entry_size (xt->n_vars));
+ te->count = weight;
+ for (j = 0; j < xt->n_vars; j++)
{
- const struct variable *var = pt->vars[j];
+ const struct variable *var = xt->vars[j].var;
value_clone (&te->values[j], case_data (c, var), var_get_width (var));
}
- hmap_insert (&pt->data, &te->node, hash);
+ hmap_insert (&xt->data, &te->node, hash);
}
\f
/* Post-data reading calculations. */
-static int compare_table_entry_vars_3way (const struct table_entry *a,
- const struct table_entry *b,
- const struct pivot_table *pt,
+static int compare_table_entry_vars_3way (const struct freq *a,
+ const struct freq *b,
+ const struct crosstabulation *xt,
int idx0, int idx1);
static int compare_table_entry_3way (const void *ap_, const void *bp_,
- const void *pt_);
+ const void *xt_);
static int compare_table_entry_3way_inv (const void *ap_, const void *bp_,
- const void *pt_);
-
-static void enum_var_values (const struct pivot_table *, int var_idx,
- union value **valuesp, int *n_values, bool descending);
-static void output_pivot_table (struct crosstabs_proc *,
- struct pivot_table *);
-static void make_pivot_table_subset (struct pivot_table *pt,
+ const void *xt_);
+
+static void enum_var_values (const struct crosstabulation *, int var_idx,
+ bool descending);
+static void free_var_values (const struct crosstabulation *, int var_idx);
+static void output_crosstabulation (struct crosstabs_proc *,
+ struct crosstabulation *);
+static void make_crosstabulation_subset (struct crosstabulation *xt,
size_t row0, size_t row1,
- struct pivot_table *subset);
+ struct crosstabulation *subset);
static void make_summary_table (struct crosstabs_proc *);
-static bool find_crosstab (struct pivot_table *, size_t *row0p, size_t *row1p);
+static bool find_crosstab (struct crosstabulation *, size_t *row0p,
+ size_t *row1p);
static void
postcalc (struct crosstabs_proc *proc)
{
- struct pivot_table *pt;
+
+ /* Round hash table entries, if requested
+
+ If this causes any of the cell counts to fall to zero, delete those
+ cells. */
+ if (proc->round_cells)
+ for (struct crosstabulation *xt = proc->pivots;
+ xt < &proc->pivots[proc->n_pivots]; xt++)
+ {
+ struct freq *e, *next;
+ HMAP_FOR_EACH_SAFE (e, next, struct freq, node, &xt->data)
+ {
+ e->count = round_weight (proc, e->count);
+ if (e->count == 0.0)
+ {
+ hmap_delete (&xt->data, &e->node);
+ free (e);
+ }
+ }
+ }
/* Convert hash tables into sorted arrays of entries. */
- for (pt = &proc->pivots[0]; pt < &proc->pivots[proc->n_pivots]; pt++)
+ for (struct crosstabulation *xt = proc->pivots;
+ xt < &proc->pivots[proc->n_pivots]; xt++)
{
- struct table_entry *e;
- size_t i;
+ struct freq *e;
- pt->n_entries = hmap_count (&pt->data);
- pt->entries = xnmalloc (pt->n_entries, sizeof *pt->entries);
- i = 0;
- HMAP_FOR_EACH (e, struct table_entry, node, &pt->data)
- pt->entries[i++] = e;
- hmap_destroy (&pt->data);
+ xt->n_entries = hmap_count (&xt->data);
+ xt->entries = xnmalloc (xt->n_entries, sizeof *xt->entries);
+ size_t i = 0;
+ HMAP_FOR_EACH (e, struct freq, node, &xt->data)
+ xt->entries[i++] = e;
+ hmap_destroy (&xt->data);
- sort (pt->entries, pt->n_entries, sizeof *pt->entries,
+ sort (xt->entries, xt->n_entries, sizeof *xt->entries,
proc->descending ? compare_table_entry_3way_inv : compare_table_entry_3way,
- pt);
+ xt);
+
}
make_summary_table (proc);
/* Output each pivot table. */
- for (pt = &proc->pivots[0]; pt < &proc->pivots[proc->n_pivots]; pt++)
+ for (struct crosstabulation *xt = proc->pivots;
+ xt < &proc->pivots[proc->n_pivots]; xt++)
{
- if (proc->pivot || pt->n_vars == 2)
- output_pivot_table (proc, pt);
+ if (proc->pivot || xt->n_vars == 2)
+ output_crosstabulation (proc, xt);
else
{
size_t row0 = 0, row1 = 0;
- while (find_crosstab (pt, &row0, &row1))
+ while (find_crosstab (xt, &row0, &row1))
{
- struct pivot_table subset;
- make_pivot_table_subset (pt, row0, row1, &subset);
- output_pivot_table (proc, &subset);
+ struct crosstabulation subset;
+ make_crosstabulation_subset (xt, row0, row1, &subset);
+ output_crosstabulation (proc, &subset);
}
}
+ if (proc->barchart)
+ {
+ const struct variable **vars = xcalloc (xt->n_vars, sizeof *vars);
+ for (size_t i = 0; i < xt->n_vars; i++)
+ vars[i] = xt->vars[i].var;
+ chart_item_submit (barchart_create (vars, xt->n_vars, _("Count"),
+ false,
+ xt->entries, xt->n_entries));
+ free (vars);
+ }
}
/* Free output and prepare for next split file. */
- for (pt = &proc->pivots[0]; pt < &proc->pivots[proc->n_pivots]; pt++)
+ for (struct crosstabulation *xt = proc->pivots;
+ xt < &proc->pivots[proc->n_pivots]; xt++)
{
- size_t i;
+ xt->missing = 0.0;
- pt->missing = 0.0;
+ /* Free the members that were allocated in this function(and the values
+ owned by the entries.
- /* Free only the members that were allocated in this
- function. The other pointer members are either both
- allocated and destroyed at a lower level (in
- output_pivot_table), or both allocated and destroyed at
- a higher level (in crs_custom_tables and free_proc,
+ The other pointer members are either both allocated and destroyed at a
+ lower level (in output_crosstabulation), or both allocated and
+ destroyed at a higher level (in crs_custom_tables and free_proc,
respectively). */
- for (i = 0; i < pt->n_entries; i++)
- free (pt->entries[i]);
- free (pt->entries);
+ for (size_t i = 0; i < xt->n_vars; i++)
+ {
+ int width = var_get_width (xt->vars[i].var);
+ if (value_needs_init (width))
+ {
+ size_t j;
+
+ for (j = 0; j < xt->n_entries; j++)
+ value_destroy (&xt->entries[j]->values[i], width);
+ }
+ }
+
+ for (size_t i = 0; i < xt->n_entries; i++)
+ free (xt->entries[i]);
+ free (xt->entries);
}
}
static void
-make_pivot_table_subset (struct pivot_table *pt, size_t row0, size_t row1,
- struct pivot_table *subset)
+make_crosstabulation_subset (struct crosstabulation *xt, size_t row0,
+ size_t row1, struct crosstabulation *subset)
{
- *subset = *pt;
- if (pt->n_vars > 2)
+ *subset = *xt;
+ if (xt->n_vars > 2)
{
- assert (pt->n_consts == 0);
- subset->missing = pt->missing;
+ assert (xt->n_consts == 0);
+ subset->missing = xt->missing;
subset->n_vars = 2;
- subset->vars = pt->vars;
- subset->n_consts = pt->n_vars - 2;
- subset->const_vars = pt->vars + 2;
- subset->const_values = &pt->entries[row0]->values[2];
+ subset->vars = xt->vars;
+
+ subset->n_consts = xt->n_vars - 2;
+ subset->const_vars = xt->vars + 2;
+ for (size_t i = 0; i < subset->n_consts; i++)
+ {
+ subset->const_vars[i].n_values = 1;
+ subset->const_vars[i].values = &xt->entries[row0]->values[2 + i];
+ }
}
- subset->entries = &pt->entries[row0];
+ subset->entries = &xt->entries[row0];
subset->n_entries = row1 - row0;
}
static int
-compare_table_entry_var_3way (const struct table_entry *a,
- const struct table_entry *b,
- const struct pivot_table *pt,
+compare_table_entry_var_3way (const struct freq *a,
+ const struct freq *b,
+ const struct crosstabulation *xt,
int idx)
{
return value_compare_3way (&a->values[idx], &b->values[idx],
- var_get_width (pt->vars[idx]));
+ var_get_width (xt->vars[idx].var));
}
static int
-compare_table_entry_vars_3way (const struct table_entry *a,
- const struct table_entry *b,
- const struct pivot_table *pt,
+compare_table_entry_vars_3way (const struct freq *a,
+ const struct freq *b,
+ const struct crosstabulation *xt,
int idx0, int idx1)
{
int i;
for (i = idx1 - 1; i >= idx0; i--)
{
- int cmp = compare_table_entry_var_3way (a, b, pt, i);
+ int cmp = compare_table_entry_var_3way (a, b, xt, i);
if (cmp != 0)
return cmp;
}
return 0;
}
-/* Compare the struct table_entry at *AP to the one at *BP and
+/* Compare the struct freq at *AP to the one at *BP and
return a strcmp()-type result. */
static int
-compare_table_entry_3way (const void *ap_, const void *bp_, const void *pt_)
+compare_table_entry_3way (const void *ap_, const void *bp_, const void *xt_)
{
- const struct table_entry *const *ap = ap_;
- const struct table_entry *const *bp = bp_;
- const struct table_entry *a = *ap;
- const struct table_entry *b = *bp;
- const struct pivot_table *pt = pt_;
+ const struct freq *const *ap = ap_;
+ const struct freq *const *bp = bp_;
+ const struct freq *a = *ap;
+ const struct freq *b = *bp;
+ const struct crosstabulation *xt = xt_;
int cmp;
- cmp = compare_table_entry_vars_3way (a, b, pt, 2, pt->n_vars);
+ cmp = compare_table_entry_vars_3way (a, b, xt, 2, xt->n_vars);
if (cmp != 0)
return cmp;
- cmp = compare_table_entry_var_3way (a, b, pt, ROW_VAR);
+ cmp = compare_table_entry_var_3way (a, b, xt, ROW_VAR);
if (cmp != 0)
return cmp;
- return compare_table_entry_var_3way (a, b, pt, COL_VAR);
+ return compare_table_entry_var_3way (a, b, xt, COL_VAR);
}
/* Inverted version of compare_table_entry_3way */
static int
-compare_table_entry_3way_inv (const void *ap_, const void *bp_, const void *pt_)
+compare_table_entry_3way_inv (const void *ap_, const void *bp_, const void *xt_)
{
- return -compare_table_entry_3way (ap_, bp_, pt_);
+ return -compare_table_entry_3way (ap_, bp_, xt_);
}
static int
-find_first_difference (const struct pivot_table *pt, size_t row)
+find_first_difference (const struct crosstabulation *xt, size_t row)
{
if (row == 0)
- return pt->n_vars - 1;
+ return xt->n_vars - 1;
else
{
- const struct table_entry *a = pt->entries[row];
- const struct table_entry *b = pt->entries[row - 1];
+ const struct freq *a = xt->entries[row];
+ const struct freq *b = xt->entries[row - 1];
int col;
- for (col = pt->n_vars - 1; col >= 0; col--)
- if (compare_table_entry_var_3way (a, b, pt, col))
+ for (col = xt->n_vars - 1; col >= 0; col--)
+ if (compare_table_entry_var_3way (a, b, xt, col))
return col;
NOT_REACHED ();
}
make_summary_table (struct crosstabs_proc *proc)
{
struct tab_table *summary;
- struct pivot_table *pt;
+ struct crosstabulation *xt;
struct string name;
int i;
summary = tab_create (7, 3 + proc->n_pivots);
+ tab_set_format (summary, RC_WEIGHT, &proc->weight_format);
tab_title (summary, _("Summary."));
tab_headers (summary, 1, 0, 3, 0);
tab_joint_text (summary, 1, 0, 6, 0, TAB_CENTER, _("Cases"));
tab_offset (summary, 0, 3);
ds_init_empty (&name);
- for (pt = &proc->pivots[0]; pt < &proc->pivots[proc->n_pivots]; pt++)
+ for (xt = &proc->pivots[0]; xt < &proc->pivots[proc->n_pivots]; xt++)
{
double valid;
double n[3];
tab_hline (summary, TAL_1, 0, 6, 0);
ds_clear (&name);
- for (i = 0; i < pt->n_vars; i++)
+ for (i = 0; i < xt->n_vars; i++)
{
if (i > 0)
ds_put_cstr (&name, " * ");
- ds_put_cstr (&name, var_to_string (pt->vars[i]));
+ ds_put_cstr (&name, var_to_string (xt->vars[i].var));
}
tab_text (summary, 0, 0, TAB_LEFT, ds_cstr (&name));
valid = 0.;
- for (i = 0; i < pt->n_entries; i++)
- valid += pt->entries[i]->freq;
+ for (i = 0; i < xt->n_entries; i++)
+ valid += xt->entries[i]->count;
n[0] = valid;
- n[1] = pt->missing;
+ n[1] = xt->missing;
n[2] = n[0] + n[1];
for (i = 0; i < 3; i++)
{
- tab_double (summary, i * 2 + 1, 0, TAB_RIGHT, n[i],
- &proc->weight_format);
+ tab_double (summary, i * 2 + 1, 0, TAB_RIGHT, n[i], NULL, RC_WEIGHT);
tab_text_format (summary, i * 2 + 2, 0, TAB_RIGHT, "%.1f%%",
n[i] / n[2] * 100.);
}
/* Output. */
static struct tab_table *create_crosstab_table (struct crosstabs_proc *,
- struct pivot_table *);
-static struct tab_table *create_chisq_table (struct pivot_table *);
-static struct tab_table *create_sym_table (struct pivot_table *);
-static struct tab_table *create_risk_table (struct pivot_table *);
-static struct tab_table *create_direct_table (struct pivot_table *);
-static void display_dimensions (struct crosstabs_proc *, struct pivot_table *,
+ struct crosstabulation *);
+static struct tab_table *create_chisq_table (struct crosstabs_proc *proc, struct crosstabulation *);
+static struct tab_table *create_sym_table (struct crosstabs_proc *proc, struct crosstabulation *);
+static struct tab_table *create_risk_table (struct crosstabs_proc *proc, struct crosstabulation *);
+static struct tab_table *create_direct_table (struct crosstabs_proc *proc, struct crosstabulation *);
+static void display_dimensions (struct crosstabs_proc *, struct crosstabulation *,
struct tab_table *, int first_difference);
static void display_crosstabulation (struct crosstabs_proc *,
- struct pivot_table *,
+ struct crosstabulation *,
struct tab_table *);
-static void display_chisq (struct pivot_table *, struct tab_table *,
+static void display_chisq (struct crosstabulation *, struct tab_table *,
bool *showed_fisher);
-static void display_symmetric (struct crosstabs_proc *, struct pivot_table *,
+static void display_symmetric (struct crosstabs_proc *, struct crosstabulation *,
struct tab_table *);
-static void display_risk (struct pivot_table *, struct tab_table *);
-static void display_directional (struct crosstabs_proc *, struct pivot_table *,
+static void display_risk (struct crosstabulation *, struct tab_table *);
+static void display_directional (struct crosstabs_proc *, struct crosstabulation *,
struct tab_table *);
static void table_value_missing (struct crosstabs_proc *proc,
struct tab_table *table, int c, int r,
unsigned char opt, const union value *v,
const struct variable *var);
-static void delete_missing (struct pivot_table *);
-static void build_matrix (struct pivot_table *);
+static void delete_missing (struct crosstabulation *);
+static void build_matrix (struct crosstabulation *);
-/* Output pivot table PT in the context of PROC. */
+/* Output pivot table XT in the context of PROC. */
static void
-output_pivot_table (struct crosstabs_proc *proc, struct pivot_table *pt)
+output_crosstabulation (struct crosstabs_proc *proc, struct crosstabulation *xt)
{
struct tab_table *table = NULL; /* Crosstabulation table. */
struct tab_table *chisq = NULL; /* Chi-square table. */
struct tab_table *direct = NULL; /* Directional measures table. */
size_t row0, row1;
- enum_var_values (pt, COL_VAR, &pt->cols, &pt->n_cols, proc->descending);
+ enum_var_values (xt, COL_VAR, proc->descending);
- if (pt->n_cols == 0)
+ if (xt->vars[COL_VAR].n_values == 0)
{
struct string vars;
int i;
- ds_init_cstr (&vars, var_to_string (pt->vars[0]));
- for (i = 1; i < pt->n_vars; i++)
- ds_put_format (&vars, " * %s", var_to_string (pt->vars[i]));
+ ds_init_cstr (&vars, var_to_string (xt->vars[0].var));
+ for (i = 1; i < xt->n_vars; i++)
+ ds_put_format (&vars, " * %s", var_to_string (xt->vars[i].var));
/* TRANSLATORS: The %s here describes a crosstabulation. It takes the
form "var1 * var2 * var3 * ...". */
ds_cstr (&vars));
ds_destroy (&vars);
+ free_var_values (xt, COL_VAR);
return;
}
if (proc->cells)
- table = create_crosstab_table (proc, pt);
+ table = create_crosstab_table (proc, xt);
if (proc->statistics & (1u << CRS_ST_CHISQ))
- chisq = create_chisq_table (pt);
+ chisq = create_chisq_table (proc, xt);
if (proc->statistics & ((1u << CRS_ST_PHI) | (1u << CRS_ST_CC)
| (1u << CRS_ST_BTAU) | (1u << CRS_ST_CTAU)
| (1u << CRS_ST_GAMMA) | (1u << CRS_ST_CORR)
| (1u << CRS_ST_KAPPA)))
- sym = create_sym_table (pt);
+ sym = create_sym_table (proc, xt);
if (proc->statistics & (1u << CRS_ST_RISK))
- risk = create_risk_table (pt);
+ risk = create_risk_table (proc, xt);
if (proc->statistics & ((1u << CRS_ST_LAMBDA) | (1u << CRS_ST_UC)
| (1u << CRS_ST_D) | (1u << CRS_ST_ETA)))
- direct = create_direct_table (pt);
+ direct = create_direct_table (proc, xt);
row0 = row1 = 0;
- while (find_crosstab (pt, &row0, &row1))
+ while (find_crosstab (xt, &row0, &row1))
{
- struct pivot_table x;
+ struct crosstabulation x;
int first_difference;
- make_pivot_table_subset (pt, row0, row1, &x);
+ make_crosstabulation_subset (xt, row0, row1, &x);
/* Find all the row variable values. */
- enum_var_values (&x, ROW_VAR, &x.rows, &x.n_rows, proc->descending);
+ enum_var_values (&x, ROW_VAR, proc->descending);
- if (size_overflow_p (xtimes (xtimes (x.n_rows, x.n_cols),
- sizeof (double))))
+ size_t n_rows = x.vars[ROW_VAR].n_values;
+ size_t n_cols = x.vars[COL_VAR].n_values;
+ if (size_overflow_p (xtimes (xtimes (n_rows, n_cols), sizeof (double))))
xalloc_die ();
- x.row_tot = xmalloc (x.n_rows * sizeof *x.row_tot);
- x.col_tot = xmalloc (x.n_cols * sizeof *x.col_tot);
- x.mat = xmalloc (x.n_rows * x.n_cols * sizeof *x.mat);
+ x.row_tot = xmalloc (n_rows * sizeof *x.row_tot);
+ x.col_tot = xmalloc (n_cols * sizeof *x.col_tot);
+ x.mat = xmalloc (n_rows * n_cols * sizeof *x.mat);
/* Allocate table space for the matrix. */
if (table
- && tab_row (table) + (x.n_rows + 1) * proc->n_cells > tab_nr (table))
+ && tab_row (table) + (n_rows + 1) * proc->n_cells > tab_nr (table))
tab_realloc (table, -1,
- MAX (tab_nr (table) + (x.n_rows + 1) * proc->n_cells,
- tab_nr (table) * pt->n_entries / x.n_entries));
+ MAX (tab_nr (table) + (n_rows + 1) * proc->n_cells,
+ tab_nr (table) * xt->n_entries / x.n_entries));
build_matrix (&x);
/* Find the first variable that differs from the last subtable. */
- first_difference = find_first_difference (pt, row0);
+ first_difference = find_first_difference (xt, row0);
if (table)
{
display_dimensions (proc, &x, table, first_difference);
display_directional (proc, &x, direct);
}
- /* Free the parts of x that are not owned by pt. In
+ /* Free the parts of x that are not owned by xt. In
particular we must not free x.cols, which is the same as
- pt->cols, which is freed at the end of this function. */
- free (x.rows);
+ xt->cols, which is freed at the end of this function. */
+ free_var_values (&x, ROW_VAR);
free (x.mat);
free (x.row_tot);
if (chisq)
{
if (!showed_fisher)
- tab_resize (chisq, 4 + (pt->n_vars - 2), -1);
- submit (pt, chisq);
+ tab_resize (chisq, 4 + (xt->n_vars - 2), -1);
+ submit (xt, chisq);
}
- submit (pt, sym);
- submit (pt, risk);
- submit (pt, direct);
+ submit (xt, sym);
+ submit (xt, risk);
+ submit (xt, direct);
- free (pt->cols);
+ free_var_values (xt, COL_VAR);
}
static void
-build_matrix (struct pivot_table *x)
+build_matrix (struct crosstabulation *x)
{
- const int col_var_width = var_get_width (x->vars[COL_VAR]);
- const int row_var_width = var_get_width (x->vars[ROW_VAR]);
+ const int col_var_width = var_get_width (x->vars[COL_VAR].var);
+ const int row_var_width = var_get_width (x->vars[ROW_VAR].var);
+ size_t n_rows = x->vars[ROW_VAR].n_values;
+ size_t n_cols = x->vars[COL_VAR].n_values;
int col, row;
double *mp;
- struct table_entry **p;
+ struct freq **p;
mp = x->mat;
col = row = 0;
for (p = x->entries; p < &x->entries[x->n_entries]; p++)
{
- const struct table_entry *te = *p;
+ const struct freq *te = *p;
- while (!value_equal (&x->rows[row], &te->values[ROW_VAR], row_var_width))
+ while (!value_equal (&x->vars[ROW_VAR].values[row],
+ &te->values[ROW_VAR], row_var_width))
{
- for (; col < x->n_cols; col++)
+ for (; col < n_cols; col++)
*mp++ = 0.0;
col = 0;
row++;
}
- while (!value_equal (&x->cols[col], &te->values[COL_VAR], col_var_width))
+ while (!value_equal (&x->vars[COL_VAR].values[col],
+ &te->values[COL_VAR], col_var_width))
{
*mp++ = 0.0;
col++;
}
- *mp++ = te->freq;
- if (++col >= x->n_cols)
+ *mp++ = te->count;
+ if (++col >= n_cols)
{
col = 0;
row++;
}
}
- while (mp < &x->mat[x->n_cols * x->n_rows])
+ while (mp < &x->mat[n_cols * n_rows])
*mp++ = 0.0;
- assert (mp == &x->mat[x->n_cols * x->n_rows]);
+ assert (mp == &x->mat[n_cols * n_rows]);
/* Column totals, row totals, ns_rows. */
mp = x->mat;
- for (col = 0; col < x->n_cols; col++)
+ for (col = 0; col < n_cols; col++)
x->col_tot[col] = 0.0;
- for (row = 0; row < x->n_rows; row++)
+ for (row = 0; row < n_rows; row++)
x->row_tot[row] = 0.0;
x->ns_rows = 0;
- for (row = 0; row < x->n_rows; row++)
+ for (row = 0; row < n_rows; row++)
{
bool row_is_empty = true;
- for (col = 0; col < x->n_cols; col++)
+ for (col = 0; col < n_cols; col++)
{
if (*mp != 0.0)
{
if (!row_is_empty)
x->ns_rows++;
}
- assert (mp == &x->mat[x->n_cols * x->n_rows]);
+ assert (mp == &x->mat[n_cols * n_rows]);
/* ns_cols. */
x->ns_cols = 0;
- for (col = 0; col < x->n_cols; col++)
- for (row = 0; row < x->n_rows; row++)
- if (x->mat[col + row * x->n_cols] != 0.0)
+ for (col = 0; col < n_cols; col++)
+ for (row = 0; row < n_rows; row++)
+ if (x->mat[col + row * n_cols] != 0.0)
{
x->ns_cols++;
break;
/* Grand total. */
x->total = 0.0;
- for (col = 0; col < x->n_cols; col++)
+ for (col = 0; col < n_cols; col++)
x->total += x->col_tot[col];
}
static struct tab_table *
-create_crosstab_table (struct crosstabs_proc *proc, struct pivot_table *pt)
+create_crosstab_table (struct crosstabs_proc *proc, struct crosstabulation *xt)
{
struct tuple
{
struct tab_table *table;
struct string title;
- struct pivot_table x;
+ struct crosstabulation x;
int i;
- make_pivot_table_subset (pt, 0, 0, &x);
+ make_crosstabulation_subset (xt, 0, 0, &x);
- table = tab_create (x.n_consts + 1 + x.n_cols + 1,
- (x.n_entries / x.n_cols) * 3 / 2 * proc->n_cells + 10);
+ size_t n_cols = x.vars[COL_VAR].n_values;
+ table = tab_create (x.n_consts + 1 + n_cols + 1,
+ (x.n_entries / n_cols) * 3 / 2 * proc->n_cells + 10);
tab_headers (table, x.n_consts + 1, 0, 2, 0);
+ tab_set_format (table, RC_WEIGHT, &proc->weight_format);
/* First header line. */
tab_joint_text (table, x.n_consts + 1, 0,
- (x.n_consts + 1) + (x.n_cols - 1), 0,
- TAB_CENTER | TAT_TITLE, var_to_string (x.vars[COL_VAR]));
+ (x.n_consts + 1) + (n_cols - 1), 0,
+ TAB_CENTER | TAT_TITLE, var_to_string (x.vars[COL_VAR].var));
tab_hline (table, TAL_1, x.n_consts + 1,
- x.n_consts + 2 + x.n_cols - 2, 1);
+ x.n_consts + 2 + n_cols - 2, 1);
/* Second header line. */
for (i = 2; i < x.n_consts + 2; i++)
tab_joint_text (table, x.n_consts + 2 - i - 1, 0,
x.n_consts + 2 - i - 1, 1,
- TAB_RIGHT | TAT_TITLE, var_to_string (x.vars[i]));
+ TAB_RIGHT | TAT_TITLE, var_to_string (x.vars[i].var));
tab_text (table, x.n_consts + 2 - 2, 1, TAB_RIGHT | TAT_TITLE,
- var_to_string (x.vars[ROW_VAR]));
- for (i = 0; i < x.n_cols; i++)
+ var_to_string (x.vars[ROW_VAR].var));
+ for (i = 0; i < n_cols; i++)
table_value_missing (proc, table, x.n_consts + 2 + i - 1, 1, TAB_RIGHT,
- &x.cols[i], x.vars[COL_VAR]);
- tab_text (table, x.n_consts + 2 + x.n_cols - 1, 1, TAB_CENTER, _("Total"));
+ &x.vars[COL_VAR].values[i], x.vars[COL_VAR].var);
+ tab_text (table, x.n_consts + 2 + n_cols - 1, 1, TAB_CENTER, _("Total"));
- tab_hline (table, TAL_1, 0, x.n_consts + 2 + x.n_cols - 1, 2);
- tab_vline (table, TAL_1, x.n_consts + 2 + x.n_cols - 1, 0, 1);
+ tab_hline (table, TAL_1, 0, x.n_consts + 2 + n_cols - 1, 2);
+ tab_vline (table, TAL_1, x.n_consts + 2 + n_cols - 1, 0, 1);
/* Title. */
ds_init_empty (&title);
{
if (i)
ds_put_cstr (&title, " * ");
- ds_put_cstr (&title, var_to_string (x.vars[i]));
+ ds_put_cstr (&title, var_to_string (x.vars[i].var));
}
- for (i = 0; i < pt->n_consts; i++)
+ for (i = 0; i < xt->n_consts; i++)
{
- const struct variable *var = pt->const_vars[i];
+ const struct variable *var = xt->const_vars[i].var;
char *s;
ds_put_format (&title, ", %s=", var_to_string (var));
/* Insert the formatted value of VAR without any leading spaces. */
- s = data_out (&pt->const_values[i], var_get_encoding (var),
+ s = data_out (&xt->const_vars[i].values[0], var_get_encoding (var),
var_get_print_format (var));
ds_put_cstr (&title, s + strspn (s, " "));
free (s);
}
static struct tab_table *
-create_chisq_table (struct pivot_table *pt)
+create_chisq_table (struct crosstabs_proc *proc, struct crosstabulation *xt)
{
struct tab_table *chisq;
- chisq = tab_create (6 + (pt->n_vars - 2),
- pt->n_entries / pt->n_cols * 3 / 2 * N_CHISQ + 10);
- tab_headers (chisq, 1 + (pt->n_vars - 2), 0, 1, 0);
+ size_t n_cols = xt->vars[COL_VAR].n_values;
+ chisq = tab_create (6 + (xt->n_vars - 2),
+ xt->n_entries / n_cols * 3 / 2 * N_CHISQ + 10);
+ tab_headers (chisq, 1 + (xt->n_vars - 2), 0, 1, 0);
+ tab_set_format (chisq, RC_WEIGHT, &proc->weight_format);
tab_title (chisq, _("Chi-square tests."));
- tab_offset (chisq, pt->n_vars - 2, 0);
+ tab_offset (chisq, xt->n_vars - 2, 0);
tab_text (chisq, 0, 0, TAB_LEFT | TAT_TITLE, _("Statistic"));
tab_text (chisq, 1, 0, TAB_RIGHT | TAT_TITLE, _("Value"));
tab_text (chisq, 2, 0, TAB_RIGHT | TAT_TITLE, _("df"));
/* Symmetric measures. */
static struct tab_table *
-create_sym_table (struct pivot_table *pt)
+create_sym_table (struct crosstabs_proc *proc, struct crosstabulation *xt)
{
struct tab_table *sym;
- sym = tab_create (6 + (pt->n_vars - 2),
- pt->n_entries / pt->n_cols * 7 + 10);
- tab_headers (sym, 2 + (pt->n_vars - 2), 0, 1, 0);
+ size_t n_cols = xt->vars[COL_VAR].n_values;
+ sym = tab_create (6 + (xt->n_vars - 2),
+ xt->n_entries / n_cols * 7 + 10);
+
+ tab_set_format (sym, RC_WEIGHT, &proc->weight_format);
+
+ tab_headers (sym, 2 + (xt->n_vars - 2), 0, 1, 0);
tab_title (sym, _("Symmetric measures."));
- tab_offset (sym, pt->n_vars - 2, 0);
+ tab_offset (sym, xt->n_vars - 2, 0);
tab_text (sym, 0, 0, TAB_LEFT | TAT_TITLE, _("Category"));
tab_text (sym, 1, 0, TAB_LEFT | TAT_TITLE, _("Statistic"));
tab_text (sym, 2, 0, TAB_RIGHT | TAT_TITLE, _("Value"));
/* Risk estimate. */
static struct tab_table *
-create_risk_table (struct pivot_table *pt)
+create_risk_table (struct crosstabs_proc *proc, struct crosstabulation *xt)
{
struct tab_table *risk;
- risk = tab_create (4 + (pt->n_vars - 2), pt->n_entries / pt->n_cols * 4 + 10);
- tab_headers (risk, 1 + pt->n_vars - 2, 0, 2, 0);
+ size_t n_cols = xt->vars[COL_VAR].n_values;
+ risk = tab_create (4 + (xt->n_vars - 2), xt->n_entries / n_cols * 4 + 10);
+ tab_headers (risk, 1 + xt->n_vars - 2, 0, 2, 0);
tab_title (risk, _("Risk estimate."));
+ tab_set_format (risk, RC_WEIGHT, &proc->weight_format);
- tab_offset (risk, pt->n_vars - 2, 0);
+ tab_offset (risk, xt->n_vars - 2, 0);
tab_joint_text_format (risk, 2, 0, 3, 0, TAB_CENTER | TAT_TITLE,
_("95%% Confidence Interval"));
tab_text (risk, 0, 1, TAB_LEFT | TAT_TITLE, _("Statistic"));
/* Directional measures. */
static struct tab_table *
-create_direct_table (struct pivot_table *pt)
+create_direct_table (struct crosstabs_proc *proc, struct crosstabulation *xt)
{
struct tab_table *direct;
- direct = tab_create (7 + (pt->n_vars - 2),
- pt->n_entries / pt->n_cols * 7 + 10);
- tab_headers (direct, 3 + (pt->n_vars - 2), 0, 1, 0);
+ size_t n_cols = xt->vars[COL_VAR].n_values;
+ direct = tab_create (7 + (xt->n_vars - 2),
+ xt->n_entries / n_cols * 7 + 10);
+ tab_headers (direct, 3 + (xt->n_vars - 2), 0, 1, 0);
tab_title (direct, _("Directional measures."));
+ tab_set_format (direct, RC_WEIGHT, &proc->weight_format);
- tab_offset (direct, pt->n_vars - 2, 0);
+ tab_offset (direct, xt->n_vars - 2, 0);
tab_text (direct, 0, 0, TAB_LEFT | TAT_TITLE, _("Category"));
tab_text (direct, 1, 0, TAB_LEFT | TAT_TITLE, _("Statistic"));
tab_text (direct, 2, 0, TAB_LEFT | TAT_TITLE, _("Type"));
/* Delete missing rows and columns for statistical analysis when
/MISSING=REPORT. */
static void
-delete_missing (struct pivot_table *pt)
+delete_missing (struct crosstabulation *xt)
{
+ size_t n_rows = xt->vars[ROW_VAR].n_values;
+ size_t n_cols = xt->vars[COL_VAR].n_values;
int r, c;
- for (r = 0; r < pt->n_rows; r++)
- if (var_is_num_missing (pt->vars[ROW_VAR], pt->rows[r].f, MV_USER))
+ for (r = 0; r < n_rows; r++)
+ if (var_is_num_missing (xt->vars[ROW_VAR].var,
+ xt->vars[ROW_VAR].values[r].f, MV_USER))
{
- for (c = 0; c < pt->n_cols; c++)
- pt->mat[c + r * pt->n_cols] = 0.;
- pt->ns_rows--;
+ for (c = 0; c < n_cols; c++)
+ xt->mat[c + r * n_cols] = 0.;
+ xt->ns_rows--;
}
- for (c = 0; c < pt->n_cols; c++)
- if (var_is_num_missing (pt->vars[COL_VAR], pt->cols[c].f, MV_USER))
+ for (c = 0; c < n_cols; c++)
+ if (var_is_num_missing (xt->vars[COL_VAR].var,
+ xt->vars[COL_VAR].values[c].f, MV_USER))
{
- for (r = 0; r < pt->n_rows; r++)
- pt->mat[c + r * pt->n_cols] = 0.;
- pt->ns_cols--;
+ for (r = 0; r < n_rows; r++)
+ xt->mat[c + r * n_cols] = 0.;
+ xt->ns_cols--;
}
}
/* Prepare table T for submission, and submit it. */
static void
-submit (struct pivot_table *pt, struct tab_table *t)
+submit (struct crosstabulation *xt, struct tab_table *t)
{
int i;
return;
}
tab_offset (t, 0, 0);
- if (pt != NULL)
- for (i = 2; i < pt->n_vars; i++)
- tab_text (t, pt->n_vars - i - 1, 0, TAB_RIGHT | TAT_TITLE,
- var_to_string (pt->vars[i]));
+ if (xt != NULL)
+ for (i = 2; i < xt->n_vars; i++)
+ tab_text (t, xt->n_vars - i - 1, 0, TAB_RIGHT | TAT_TITLE,
+ var_to_string (xt->vars[i].var));
tab_box (t, TAL_2, TAL_2, -1, -1, 0, 0, tab_nc (t) - 1, tab_nr (t) - 1);
tab_box (t, -1, -1, -1, TAL_1, tab_l (t), tab_t (t) - 1, tab_nc (t) - 1,
tab_nr (t) - 1);
- tab_box (t, -1, -1, -1, TAL_GAP, 0, tab_t (t), tab_l (t) - 1,
- tab_nr (t) - 1);
tab_vline (t, TAL_2, tab_l (t), 0, tab_nr (t) - 1);
tab_submit (t);
}
static bool
-find_crosstab (struct pivot_table *pt, size_t *row0p, size_t *row1p)
+find_crosstab (struct crosstabulation *xt, size_t *row0p, size_t *row1p)
{
size_t row0 = *row1p;
size_t row1;
- if (row0 >= pt->n_entries)
+ if (row0 >= xt->n_entries)
return false;
- for (row1 = row0 + 1; row1 < pt->n_entries; row1++)
+ for (row1 = row0 + 1; row1 < xt->n_entries; row1++)
{
- struct table_entry *a = pt->entries[row0];
- struct table_entry *b = pt->entries[row1];
- if (compare_table_entry_vars_3way (a, b, pt, 2, pt->n_vars) != 0)
+ struct freq *a = xt->entries[row0];
+ struct freq *b = xt->entries[row1];
+ if (compare_table_entry_vars_3way (a, b, xt, 2, xt->n_vars) != 0)
break;
}
*row0p = row0;
with index VAR_IDX takes on. The values are returned as a
malloc()'d array stored in *VALUES, with the number of values
stored in *VALUE_CNT.
- */
+
+ The caller must eventually free *VALUES, but each pointer in *VALUES points
+ to existing data not owned by *VALUES itself. */
static void
-enum_var_values (const struct pivot_table *pt, int var_idx,
- union value **valuesp, int *n_values, bool descending)
+enum_var_values (const struct crosstabulation *xt, int var_idx,
+ bool descending)
{
- const struct variable *var = pt->vars[var_idx];
- struct var_range *range = get_var_range (var);
- union value *values;
- size_t i;
+ struct xtab_var *xv = &xt->vars[var_idx];
+ const struct var_range *range = get_var_range (xt->proc, xv->var);
if (range)
{
- values = *valuesp = xnmalloc (range->count, sizeof *values);
- *n_values = range->count;
- for (i = 0; i < range->count; i++)
- values[i].f = range->min + i;
+ xv->values = xnmalloc (range->count, sizeof *xv->values);
+ xv->n_values = range->count;
+ for (size_t i = 0; i < range->count; i++)
+ xv->values[i].f = range->min + i;
}
else
{
- int width = var_get_width (var);
+ int width = var_get_width (xv->var);
struct hmapx_node *node;
const union value *iter;
struct hmapx set;
hmapx_init (&set);
- for (i = 0; i < pt->n_entries; i++)
+ for (size_t i = 0; i < xt->n_entries; i++)
{
- const struct table_entry *te = pt->entries[i];
+ const struct freq *te = xt->entries[i];
const union value *value = &te->values[var_idx];
size_t hash = value_hash (value, width, 0);
next_entry: ;
}
- *n_values = hmapx_count (&set);
- values = *valuesp = xnmalloc (*n_values, sizeof *values);
- i = 0;
+ xv->n_values = hmapx_count (&set);
+ xv->values = xnmalloc (xv->n_values, sizeof *xv->values);
+ size_t i = 0;
HMAPX_FOR_EACH (iter, node, &set)
- values[i++] = *iter;
+ xv->values[i++] = *iter;
hmapx_destroy (&set);
- sort (values, *n_values, sizeof *values,
+ sort (xv->values, xv->n_values, sizeof *xv->values,
descending ? compare_value_3way_inv : compare_value_3way,
&width);
}
}
+static void
+free_var_values (const struct crosstabulation *xt, int var_idx)
+{
+ struct xtab_var *xv = &xt->vars[var_idx];
+ free (xv->values);
+ xv->values = NULL;
+ xv->n_values = 0;
+}
+
/* Sets cell (C,R) in TABLE, with options OPT, to have a value taken
from V, displayed with print format spec from variable VAR. When
in REPORT missing-value mode, missing values have an M appended. */
/* Draws a line across TABLE at the current row to indicate the most
major dimension variable with index FIRST_DIFFERENCE out of N_VARS
that changed, and puts the values that changed into the table. TB
- and PT must be the corresponding table_entry and crosstab,
+ and XT must be the corresponding table_entry and crosstab,
respectively. */
static void
-display_dimensions (struct crosstabs_proc *proc, struct pivot_table *pt,
+display_dimensions (struct crosstabs_proc *proc, struct crosstabulation *xt,
struct tab_table *table, int first_difference)
{
- tab_hline (table, TAL_1, pt->n_consts + pt->n_vars - first_difference - 1, tab_nc (table) - 1, 0);
+ tab_hline (table, TAL_1, xt->n_consts + xt->n_vars - first_difference - 1, tab_nc (table) - 1, 0);
for (; first_difference >= 2; first_difference--)
- table_value_missing (proc, table, pt->n_consts + pt->n_vars - first_difference - 1, 0,
- TAB_RIGHT, &pt->entries[0]->values[first_difference],
- pt->vars[first_difference]);
+ table_value_missing (proc, table, xt->n_consts + xt->n_vars - first_difference - 1, 0,
+ TAB_RIGHT, &xt->entries[0]->values[first_difference],
+ xt->vars[first_difference].var);
}
/* Put VALUE into cell (C,R) of TABLE, suffixed with character
format_cell_entry (struct tab_table *table, int c, int r, double value,
char suffix, bool mark_missing, const struct dictionary *dict)
{
- const struct fmt_spec f = {FMT_F, 10, 1};
union value v;
char suffixes[3];
int suffix_len;
char *s;
v.f = value;
- s = data_out (&v, dict_get_encoding (dict), &f);
+ s = data_out (&v, dict_get_encoding (dict), settings_get_format ());
suffix_len = 0;
if (suffix != 0)
/* Displays the crosstabulation table. */
static void
-display_crosstabulation (struct crosstabs_proc *proc, struct pivot_table *pt,
- struct tab_table *table)
+display_crosstabulation (struct crosstabs_proc *proc,
+ struct crosstabulation *xt, struct tab_table *table)
{
+ size_t n_rows = xt->vars[ROW_VAR].n_values;
+ size_t n_cols = xt->vars[COL_VAR].n_values;
int last_row;
int r, c, i;
double *mp;
- for (r = 0; r < pt->n_rows; r++)
- table_value_missing (proc, table, pt->n_consts + pt->n_vars - 2,
- r * proc->n_cells, TAB_RIGHT, &pt->rows[r],
- pt->vars[ROW_VAR]);
+ for (r = 0; r < n_rows; r++)
+ table_value_missing (proc, table, xt->n_consts + xt->n_vars - 2,
+ r * proc->n_cells, TAB_RIGHT,
+ &xt->vars[ROW_VAR].values[r],
+ xt->vars[ROW_VAR].var);
- tab_text (table, pt->n_vars - 2, pt->n_rows * proc->n_cells,
+ tab_text (table, xt->n_vars - 2, n_rows * proc->n_cells,
TAB_LEFT, _("Total"));
/* Put in the actual cells. */
- mp = pt->mat;
- tab_offset (table, pt->n_consts + pt->n_vars - 1, -1);
- for (r = 0; r < pt->n_rows; r++)
+ mp = xt->mat;
+ tab_offset (table, xt->n_consts + xt->n_vars - 1, -1);
+ for (r = 0; r < n_rows; r++)
{
if (proc->n_cells > 1)
- tab_hline (table, TAL_1, -1, pt->n_cols, 0);
- for (c = 0; c < pt->n_cols; c++)
+ tab_hline (table, TAL_1, -1, n_cols, 0);
+ for (c = 0; c < n_cols; c++)
{
bool mark_missing = false;
- double expected_value = pt->row_tot[r] * pt->col_tot[c] / pt->total;
+ double expected_value = xt->row_tot[r] * xt->col_tot[c] / xt->total;
if (proc->exclude == MV_NEVER
- && (var_is_num_missing (pt->vars[COL_VAR], pt->cols[c].f, MV_USER)
- || var_is_num_missing (pt->vars[ROW_VAR], pt->rows[r].f,
+ && (var_is_num_missing (xt->vars[COL_VAR].var,
+ xt->vars[COL_VAR].values[c].f, MV_USER)
+ || var_is_num_missing (xt->vars[ROW_VAR].var,
+ xt->vars[ROW_VAR].values[r].f,
MV_USER)))
mark_missing = true;
for (i = 0; i < proc->n_cells; i++)
v = *mp;
break;
case CRS_CL_ROW:
- v = *mp / pt->row_tot[r] * 100.;
+ v = *mp / xt->row_tot[r] * 100.;
suffix = '%';
break;
case CRS_CL_COLUMN:
- v = *mp / pt->col_tot[c] * 100.;
+ v = *mp / xt->col_tot[c] * 100.;
suffix = '%';
break;
case CRS_CL_TOTAL:
- v = *mp / pt->total * 100.;
+ v = *mp / xt->total * 100.;
suffix = '%';
break;
case CRS_CL_EXPECTED:
case CRS_CL_ASRESIDUAL:
v = ((*mp - expected_value)
/ sqrt (expected_value
- * (1. - pt->row_tot[r] / pt->total)
- * (1. - pt->col_tot[c] / pt->total)));
+ * (1. - xt->row_tot[r] / xt->total)
+ * (1. - xt->col_tot[c] / xt->total)));
break;
default:
NOT_REACHED ();
}
/* Row totals. */
- tab_offset (table, -1, tab_row (table) - proc->n_cells * pt->n_rows);
- for (r = 0; r < pt->n_rows; r++)
+ tab_offset (table, -1, tab_row (table) - proc->n_cells * n_rows);
+ for (r = 0; r < n_rows; r++)
{
bool mark_missing = false;
if (proc->exclude == MV_NEVER
- && var_is_num_missing (pt->vars[ROW_VAR], pt->rows[r].f, MV_USER))
+ && var_is_num_missing (xt->vars[ROW_VAR].var,
+ xt->vars[ROW_VAR].values[r].f, MV_USER))
mark_missing = true;
for (i = 0; i < proc->n_cells; i++)
switch (proc->a_cells[i])
{
case CRS_CL_COUNT:
- v = pt->row_tot[r];
+ v = xt->row_tot[r];
break;
case CRS_CL_ROW:
v = 100.0;
suffix = '%';
break;
case CRS_CL_COLUMN:
- v = pt->row_tot[r] / pt->total * 100.;
+ v = xt->row_tot[r] / xt->total * 100.;
suffix = '%';
break;
case CRS_CL_TOTAL:
- v = pt->row_tot[r] / pt->total * 100.;
+ v = xt->row_tot[r] / xt->total * 100.;
suffix = '%';
break;
case CRS_CL_EXPECTED:
NOT_REACHED ();
}
- format_cell_entry (table, pt->n_cols, 0, v, suffix, mark_missing, proc->dict);
+ format_cell_entry (table, n_cols, 0, v, suffix, mark_missing, proc->dict);
tab_next_row (table);
}
}
/* Column totals, grand total. */
last_row = 0;
if (proc->n_cells > 1)
- tab_hline (table, TAL_1, -1, pt->n_cols, 0);
- for (c = 0; c <= pt->n_cols; c++)
+ tab_hline (table, TAL_1, -1, n_cols, 0);
+ for (c = 0; c <= n_cols; c++)
{
- double ct = c < pt->n_cols ? pt->col_tot[c] : pt->total;
+ double ct = c < n_cols ? xt->col_tot[c] : xt->total;
bool mark_missing = false;
int i;
- if (proc->exclude == MV_NEVER && c < pt->n_cols
- && var_is_num_missing (pt->vars[COL_VAR], pt->cols[c].f, MV_USER))
+ if (proc->exclude == MV_NEVER && c < n_cols
+ && var_is_num_missing (xt->vars[COL_VAR].var,
+ xt->vars[COL_VAR].values[c].f, MV_USER))
mark_missing = true;
for (i = 0; i < proc->n_cells; i++)
v = ct;
break;
case CRS_CL_ROW:
- v = ct / pt->total * 100.;
+ v = ct / xt->total * 100.;
suffix = '%';
break;
case CRS_CL_COLUMN:
suffix = '%';
break;
case CRS_CL_TOTAL:
- v = ct / pt->total * 100.;
+ v = ct / xt->total * 100.;
suffix = '%';
break;
case CRS_CL_EXPECTED:
tab_offset (table, 0, -1);
}
-static void calc_r (struct pivot_table *,
- double *PT, double *Y, double *, double *, double *);
-static void calc_chisq (struct pivot_table *,
+static void calc_r (struct crosstabulation *,
+ double *XT, double *Y, double *, double *, double *);
+static void calc_chisq (struct crosstabulation *,
double[N_CHISQ], int[N_CHISQ], double *, double *);
/* Display chi-square statistics. */
static void
-display_chisq (struct pivot_table *pt, struct tab_table *chisq,
+display_chisq (struct crosstabulation *xt, struct tab_table *chisq,
bool *showed_fisher)
{
static const char *chisq_stats[N_CHISQ] =
int i;
- calc_chisq (pt, chisq_v, df, &fisher1, &fisher2);
+ calc_chisq (xt, chisq_v, df, &fisher1, &fisher2);
- tab_offset (chisq, pt->n_consts + pt->n_vars - 2, -1);
+ tab_offset (chisq, xt->n_consts + xt->n_vars - 2, -1);
for (i = 0; i < N_CHISQ; i++)
{
tab_text (chisq, 0, 0, TAB_LEFT, gettext (chisq_stats[i]));
if (i != 2)
{
- tab_double (chisq, 1, 0, TAB_RIGHT, chisq_v[i], NULL);
- tab_double (chisq, 2, 0, TAB_RIGHT, df[i], &pt->weight_format);
+ tab_double (chisq, 1, 0, TAB_RIGHT, chisq_v[i], NULL, RC_OTHER);
+ tab_double (chisq, 2, 0, TAB_RIGHT, df[i], NULL, RC_WEIGHT);
tab_double (chisq, 3, 0, TAB_RIGHT,
- gsl_cdf_chisq_Q (chisq_v[i], df[i]), NULL);
+ gsl_cdf_chisq_Q (chisq_v[i], df[i]), NULL, RC_PVALUE);
}
else
{
*showed_fisher = true;
- tab_double (chisq, 4, 0, TAB_RIGHT, fisher2, NULL);
- tab_double (chisq, 5, 0, TAB_RIGHT, fisher1, NULL);
+ tab_double (chisq, 4, 0, TAB_RIGHT, fisher2, NULL, RC_PVALUE);
+ tab_double (chisq, 5, 0, TAB_RIGHT, fisher1, NULL, RC_PVALUE);
}
tab_next_row (chisq);
}
tab_text (chisq, 0, 0, TAB_LEFT, _("N of Valid Cases"));
- tab_double (chisq, 1, 0, TAB_RIGHT, pt->total, &pt->weight_format);
+ tab_double (chisq, 1, 0, TAB_RIGHT, xt->total, NULL, RC_WEIGHT);
tab_next_row (chisq);
tab_offset (chisq, 0, -1);
}
-static int calc_symmetric (struct crosstabs_proc *, struct pivot_table *,
+static int calc_symmetric (struct crosstabs_proc *, struct crosstabulation *,
double[N_SYMMETRIC], double[N_SYMMETRIC],
double[N_SYMMETRIC],
double[3], double[3], double[3]);
/* Display symmetric measures. */
static void
-display_symmetric (struct crosstabs_proc *proc, struct pivot_table *pt,
+display_symmetric (struct crosstabs_proc *proc, struct crosstabulation *xt,
struct tab_table *sym)
{
static const char *categories[] =
double somers_d_v[3], somers_d_ase[3], somers_d_t[3];
int i;
- if (!calc_symmetric (proc, pt, sym_v, sym_ase, sym_t,
+ if (!calc_symmetric (proc, xt, sym_v, sym_ase, sym_t,
somers_d_v, somers_d_ase, somers_d_t))
return;
- tab_offset (sym, pt->n_consts + pt->n_vars - 2, -1);
+ tab_offset (sym, xt->n_consts + xt->n_vars - 2, -1);
for (i = 0; i < N_SYMMETRIC; i++)
{
}
tab_text (sym, 1, 0, TAB_LEFT, gettext (stats[i]));
- tab_double (sym, 2, 0, TAB_RIGHT, sym_v[i], NULL);
+ tab_double (sym, 2, 0, TAB_RIGHT, sym_v[i], NULL, RC_OTHER);
if (sym_ase[i] != SYSMIS)
- tab_double (sym, 3, 0, TAB_RIGHT, sym_ase[i], NULL);
+ tab_double (sym, 3, 0, TAB_RIGHT, sym_ase[i], NULL, RC_OTHER);
if (sym_t[i] != SYSMIS)
- tab_double (sym, 4, 0, TAB_RIGHT, sym_t[i], NULL);
- /*tab_double (sym, 5, 0, TAB_RIGHT, normal_sig (sym_v[i]), NULL);*/
+ tab_double (sym, 4, 0, TAB_RIGHT, sym_t[i], NULL, RC_OTHER);
+ /*tab_double (sym, 5, 0, TAB_RIGHT, normal_sig (sym_v[i]), NULL, RC_PVALUE);*/
tab_next_row (sym);
}
tab_text (sym, 0, 0, TAB_LEFT, _("N of Valid Cases"));
- tab_double (sym, 2, 0, TAB_RIGHT, pt->total, &pt->weight_format);
+ tab_double (sym, 2, 0, TAB_RIGHT, xt->total, NULL, RC_WEIGHT);
tab_next_row (sym);
tab_offset (sym, 0, -1);
}
-static int calc_risk (struct pivot_table *,
+static int calc_risk (struct crosstabulation *,
double[], double[], double[], union value *);
/* Display risk estimate. */
static void
-display_risk (struct pivot_table *pt, struct tab_table *risk)
+display_risk (struct crosstabulation *xt, struct tab_table *risk)
{
char buf[256];
double risk_v[3], lower[3], upper[3];
union value c[2];
int i;
- if (!calc_risk (pt, risk_v, upper, lower, c))
+ if (!calc_risk (xt, risk_v, upper, lower, c))
return;
- tab_offset (risk, pt->n_consts + pt->n_vars - 2, -1);
+ tab_offset (risk, xt->n_consts + xt->n_vars - 2, -1);
for (i = 0; i < 3; i++)
{
- const struct variable *cv = pt->vars[COL_VAR];
- const struct variable *rv = pt->vars[ROW_VAR];
+ const struct variable *cv = xt->vars[COL_VAR].var;
+ const struct variable *rv = xt->vars[ROW_VAR].var;
int cvw = var_get_width (cv);
int rvw = var_get_width (rv);
case 1:
case 2:
if (var_is_numeric (rv))
- sprintf (buf, _("For cohort %s = %g"),
- var_to_string (rv), pt->rows[i - 1].f);
+ sprintf (buf, _("For cohort %s = %.*g"),
+ var_to_string (rv), DBL_DIG + 1,
+ xt->vars[ROW_VAR].values[i - 1].f);
else
sprintf (buf, _("For cohort %s = %.*s"),
var_to_string (rv),
- rvw, value_str (&pt->rows[i - 1], rvw));
+ rvw, value_str (&xt->vars[ROW_VAR].values[i - 1], rvw));
break;
}
tab_text (risk, 0, 0, TAB_LEFT, buf);
- tab_double (risk, 1, 0, TAB_RIGHT, risk_v[i], NULL);
- tab_double (risk, 2, 0, TAB_RIGHT, lower[i], NULL);
- tab_double (risk, 3, 0, TAB_RIGHT, upper[i], NULL);
+ tab_double (risk, 1, 0, TAB_RIGHT, risk_v[i], NULL, RC_OTHER);
+ tab_double (risk, 2, 0, TAB_RIGHT, lower[i], NULL, RC_OTHER);
+ tab_double (risk, 3, 0, TAB_RIGHT, upper[i], NULL, RC_OTHER);
tab_next_row (risk);
}
tab_text (risk, 0, 0, TAB_LEFT, _("N of Valid Cases"));
- tab_double (risk, 1, 0, TAB_RIGHT, pt->total, &pt->weight_format);
+ tab_double (risk, 1, 0, TAB_RIGHT, xt->total, NULL, RC_WEIGHT);
tab_next_row (risk);
tab_offset (risk, 0, -1);
}
-static int calc_directional (struct crosstabs_proc *, struct pivot_table *,
+static int calc_directional (struct crosstabs_proc *, struct crosstabulation *,
double[N_DIRECTIONAL], double[N_DIRECTIONAL],
- double[N_DIRECTIONAL]);
+ double[N_DIRECTIONAL], double[N_DIRECTIONAL]);
/* Display directional measures. */
static void
-display_directional (struct crosstabs_proc *proc, struct pivot_table *pt,
+display_directional (struct crosstabs_proc *proc, struct crosstabulation *xt,
struct tab_table *direct)
{
static const char *categories[] =
double direct_v[N_DIRECTIONAL];
double direct_ase[N_DIRECTIONAL];
double direct_t[N_DIRECTIONAL];
+ double sig[N_DIRECTIONAL];
int i;
- if (!calc_directional (proc, pt, direct_v, direct_ase, direct_t))
+ if (!calc_directional (proc, xt, direct_v, direct_ase, direct_t, sig))
return;
- tab_offset (direct, pt->n_consts + pt->n_vars - 2, -1);
+ tab_offset (direct, xt->n_consts + xt->n_vars - 2, -1);
for (i = 0; i < N_DIRECTIONAL; i++)
{
if (k == 0)
string = NULL;
else if (k == 1)
- string = var_to_string (pt->vars[0]);
+ string = var_to_string (xt->vars[0].var);
else
- string = var_to_string (pt->vars[1]);
+ string = var_to_string (xt->vars[1].var);
tab_text_format (direct, j, 0, TAB_LEFT,
gettext (stats_names[j][k]), string);
}
}
- tab_double (direct, 3, 0, TAB_RIGHT, direct_v[i], NULL);
+ tab_double (direct, 3, 0, TAB_RIGHT, direct_v[i], NULL, RC_OTHER);
if (direct_ase[i] != SYSMIS)
- tab_double (direct, 4, 0, TAB_RIGHT, direct_ase[i], NULL);
+ tab_double (direct, 4, 0, TAB_RIGHT, direct_ase[i], NULL, RC_OTHER);
if (direct_t[i] != SYSMIS)
- tab_double (direct, 5, 0, TAB_RIGHT, direct_t[i], NULL);
- /*tab_double (direct, 6, 0, TAB_RIGHT, normal_sig (direct_v[i]), NULL);*/
+ tab_double (direct, 5, 0, TAB_RIGHT, direct_t[i], NULL, RC_OTHER);
+ tab_double (direct, 6, 0, TAB_RIGHT, sig[i], NULL, RC_PVALUE);
tab_next_row (direct);
}
\f
/* Statistical calculations. */
-/* Returns the value of the gamma (factorial) function for an integer
- argument PT. */
+/* Returns the value of the logarithm of gamma (factorial) function for an integer
+ argument XT. */
static double
-gamma_int (double pt)
+log_gamma_int (double xt)
{
- double r = 1;
+ double r = 0;
int i;
- for (i = 2; i < pt; i++)
- r *= i;
+ for (i = 2; i < xt; i++)
+ r += log(i);
+
return r;
}
static inline double
Pr (int a, int b, int c, int d)
{
- return (gamma_int (a + b + 1.) / gamma_int (a + 1.)
- * gamma_int (c + d + 1.) / gamma_int (b + 1.)
- * gamma_int (a + c + 1.) / gamma_int (c + 1.)
- * gamma_int (b + d + 1.) / gamma_int (d + 1.)
- / gamma_int (a + b + c + d + 1.));
+ return exp (log_gamma_int (a + b + 1.) - log_gamma_int (a + 1.)
+ + log_gamma_int (c + d + 1.) - log_gamma_int (b + 1.)
+ + log_gamma_int (a + c + 1.) - log_gamma_int (c + 1.)
+ + log_gamma_int (b + d + 1.) - log_gamma_int (d + 1.)
+ - log_gamma_int (a + b + c + d + 1.));
}
/* Swap the contents of A and B. */
static void
calc_fisher (int a, int b, int c, int d, double *fisher1, double *fisher2)
{
- int pt;
+ int xt;
+ double pn1;
if (MIN (c, d) < MIN (a, b))
swap (&a, &c), swap (&b, &d);
swap (&a, &c), swap (&b, &d);
}
- *fisher1 = 0.;
- for (pt = 0; pt <= a; pt++)
- *fisher1 += Pr (a - pt, b + pt, c + pt, d - pt);
+ pn1 = Pr (a, b, c, d);
+ *fisher1 = pn1;
+ for (xt = 1; xt <= a; xt++)
+ {
+ *fisher1 += Pr (a - xt, b + xt, c + xt, d - xt);
+ }
*fisher2 = *fisher1;
- for (pt = 1; pt <= b; pt++)
- *fisher2 += Pr (a + pt, b - pt, c - pt, d + pt);
+
+ for (xt = 1; xt <= b; xt++)
+ {
+ double p = Pr (a + xt, b - xt, c - xt, d + xt);
+ if (p < pn1)
+ *fisher2 += p;
+ }
}
/* Calculates chi-squares into CHISQ. MAT is a matrix with N_COLS
columns with values COLS and N_ROWS rows with values ROWS. Values
- in the matrix sum to pt->total. */
+ in the matrix sum to xt->total. */
static void
-calc_chisq (struct pivot_table *pt,
+calc_chisq (struct crosstabulation *xt,
double chisq[N_CHISQ], int df[N_CHISQ],
double *fisher1, double *fisher2)
{
chisq[2] = chisq[3] = chisq[4] = SYSMIS;
*fisher1 = *fisher2 = SYSMIS;
- df[0] = df[1] = (pt->ns_cols - 1) * (pt->ns_rows - 1);
+ df[0] = df[1] = (xt->ns_cols - 1) * (xt->ns_rows - 1);
- if (pt->ns_rows <= 1 || pt->ns_cols <= 1)
+ if (xt->ns_rows <= 1 || xt->ns_cols <= 1)
{
chisq[0] = chisq[1] = SYSMIS;
return;
}
- for (r = 0; r < pt->n_rows; r++)
- for (c = 0; c < pt->n_cols; c++)
+ size_t n_rows = xt->vars[ROW_VAR].n_values;
+ size_t n_cols = xt->vars[COL_VAR].n_values;
+ for (r = 0; r < n_rows; r++)
+ for (c = 0; c < n_cols; c++)
{
- const double expected = pt->row_tot[r] * pt->col_tot[c] / pt->total;
- const double freq = pt->mat[pt->n_cols * r + c];
+ const double expected = xt->row_tot[r] * xt->col_tot[c] / xt->total;
+ const double freq = xt->mat[n_cols * r + c];
const double residual = freq - expected;
chisq[0] += residual * residual / expected;
chisq[1] = SYSMIS;
/* Calculate Yates and Fisher exact test. */
- if (pt->ns_cols == 2 && pt->ns_rows == 2)
+ if (xt->ns_cols == 2 && xt->ns_rows == 2)
{
double f11, f12, f21, f22;
int nz_cols[2];
int i, j;
- for (i = j = 0; i < pt->n_cols; i++)
- if (pt->col_tot[i] != 0.)
+ for (i = j = 0; i < n_cols; i++)
+ if (xt->col_tot[i] != 0.)
{
nz_cols[j++] = i;
if (j == 2)
assert (j == 2);
- f11 = pt->mat[nz_cols[0]];
- f12 = pt->mat[nz_cols[1]];
- f21 = pt->mat[nz_cols[0] + pt->n_cols];
- f22 = pt->mat[nz_cols[1] + pt->n_cols];
+ f11 = xt->mat[nz_cols[0]];
+ f12 = xt->mat[nz_cols[1]];
+ f21 = xt->mat[nz_cols[0] + n_cols];
+ f22 = xt->mat[nz_cols[1] + n_cols];
}
/* Yates. */
{
- const double pt_ = fabs (f11 * f22 - f12 * f21) - 0.5 * pt->total;
+ const double xt_ = fabs (f11 * f22 - f12 * f21) - 0.5 * xt->total;
- if (pt_ > 0.)
- chisq[3] = (pt->total * pow2 (pt_)
+ if (xt_ > 0.)
+ chisq[3] = (xt->total * pow2 (xt_)
/ (f11 + f12) / (f21 + f22)
/ (f11 + f21) / (f12 + f22));
else
}
/* Fisher. */
- if (f11 < 5. || f12 < 5. || f21 < 5. || f22 < 5.)
- calc_fisher (f11 + .5, f12 + .5, f21 + .5, f22 + .5, fisher1, fisher2);
+ calc_fisher (f11 + .5, f12 + .5, f21 + .5, f22 + .5, fisher1, fisher2);
}
/* Calculate Mantel-Haenszel. */
- if (var_is_numeric (pt->vars[ROW_VAR]) && var_is_numeric (pt->vars[COL_VAR]))
+ if (var_is_numeric (xt->vars[ROW_VAR].var)
+ && var_is_numeric (xt->vars[COL_VAR].var))
{
double r, ase_0, ase_1;
- calc_r (pt, (double *) pt->rows, (double *) pt->cols, &r, &ase_0, &ase_1);
+ calc_r (xt, (double *) xt->vars[ROW_VAR].values,
+ (double *) xt->vars[COL_VAR].values,
+ &r, &ase_0, &ase_1);
- chisq[4] = (pt->total - 1.) * r * r;
+ chisq[4] = (xt->total - 1.) * r * r;
df[4] = 1;
}
}
-/* Calculate the value of Pearson's r. r is stored into R, ase_1 into
- ASE_1, and ase_0 into ASE_0. The row and column values must be
- passed in PT and Y. */
+/* Calculate the value of Pearson's r. r is stored into R, its T value into
+ T, and standard error into ERROR. The row and column values must be
+ passed in XT and Y. */
static void
-calc_r (struct pivot_table *pt,
- double *PT, double *Y, double *r, double *ase_0, double *ase_1)
+calc_r (struct crosstabulation *xt,
+ double *XT, double *Y, double *r, double *t, double *error)
{
+ size_t n_rows = xt->vars[ROW_VAR].n_values;
+ size_t n_cols = xt->vars[COL_VAR].n_values;
double SX, SY, S, T;
double Xbar, Ybar;
double sum_XYf, sum_X2Y2f;
double sum_Yc, sum_Y2c;
int i, j;
- for (sum_X2Y2f = sum_XYf = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
+ for (sum_X2Y2f = sum_XYf = 0., i = 0; i < n_rows; i++)
+ for (j = 0; j < n_cols; j++)
{
- double fij = pt->mat[j + i * pt->n_cols];
- double product = PT[i] * Y[j];
+ double fij = xt->mat[j + i * n_cols];
+ double product = XT[i] * Y[j];
double temp = fij * product;
sum_XYf += temp;
sum_X2Y2f += temp * product;
}
- for (sum_Xr = sum_X2r = 0., i = 0; i < pt->n_rows; i++)
+ for (sum_Xr = sum_X2r = 0., i = 0; i < n_rows; i++)
{
- sum_Xr += PT[i] * pt->row_tot[i];
- sum_X2r += pow2 (PT[i]) * pt->row_tot[i];
+ sum_Xr += XT[i] * xt->row_tot[i];
+ sum_X2r += pow2 (XT[i]) * xt->row_tot[i];
}
- Xbar = sum_Xr / pt->total;
+ Xbar = sum_Xr / xt->total;
- for (sum_Yc = sum_Y2c = 0., i = 0; i < pt->n_cols; i++)
+ for (sum_Yc = sum_Y2c = 0., i = 0; i < n_cols; i++)
{
- sum_Yc += Y[i] * pt->col_tot[i];
- sum_Y2c += Y[i] * Y[i] * pt->col_tot[i];
+ sum_Yc += Y[i] * xt->col_tot[i];
+ sum_Y2c += Y[i] * Y[i] * xt->col_tot[i];
}
- Ybar = sum_Yc / pt->total;
+ Ybar = sum_Yc / xt->total;
- S = sum_XYf - sum_Xr * sum_Yc / pt->total;
- SX = sum_X2r - pow2 (sum_Xr) / pt->total;
- SY = sum_Y2c - pow2 (sum_Yc) / pt->total;
+ S = sum_XYf - sum_Xr * sum_Yc / xt->total;
+ SX = sum_X2r - pow2 (sum_Xr) / xt->total;
+ SY = sum_Y2c - pow2 (sum_Yc) / xt->total;
T = sqrt (SX * SY);
*r = S / T;
- *ase_0 = sqrt ((sum_X2Y2f - pow2 (sum_XYf) / pt->total) / (sum_X2r * sum_Y2c));
+ *t = *r / sqrt (1 - pow2 (*r)) * sqrt (xt->total - 2);
{
double s, c, y, t;
- for (s = c = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
+ for (s = c = 0., i = 0; i < n_rows; i++)
+ for (j = 0; j < n_cols; j++)
{
double Xresid, Yresid;
double temp;
- Xresid = PT[i] - Xbar;
+ Xresid = XT[i] - Xbar;
Yresid = Y[j] - Ybar;
temp = (T * Xresid * Yresid
- ((S / (2. * T))
* (Xresid * Xresid * SY + Yresid * Yresid * SX)));
- y = pt->mat[j + i * pt->n_cols] * temp * temp - c;
+ y = xt->mat[j + i * n_cols] * temp * temp - c;
t = s + y;
c = (t - s) - y;
s = t;
}
- *ase_1 = sqrt (s) / (T * T);
+ *error = sqrt (s) / (T * T);
}
}
/* Calculate symmetric statistics and their asymptotic standard
errors. Returns 0 if none could be calculated. */
static int
-calc_symmetric (struct crosstabs_proc *proc, struct pivot_table *pt,
+calc_symmetric (struct crosstabs_proc *proc, struct crosstabulation *xt,
double v[N_SYMMETRIC], double ase[N_SYMMETRIC],
double t[N_SYMMETRIC],
double somers_d_v[3], double somers_d_ase[3],
double somers_d_t[3])
{
+ size_t n_rows = xt->vars[ROW_VAR].n_values;
+ size_t n_cols = xt->vars[COL_VAR].n_values;
int q, i;
- q = MIN (pt->ns_rows, pt->ns_cols);
+ q = MIN (xt->ns_rows, xt->ns_cols);
if (q <= 1)
return 0;
double Xp = 0.; /* Pearson chi-square. */
int r, c;
- for (r = 0; r < pt->n_rows; r++)
- for (c = 0; c < pt->n_cols; c++)
+ for (r = 0; r < n_rows; r++)
+ for (c = 0; c < n_cols; c++)
{
- const double expected = pt->row_tot[r] * pt->col_tot[c] / pt->total;
- const double freq = pt->mat[pt->n_cols * r + c];
+ const double expected = xt->row_tot[r] * xt->col_tot[c] / xt->total;
+ const double freq = xt->mat[n_cols * r + c];
const double residual = freq - expected;
Xp += residual * residual / expected;
if (proc->statistics & (1u << CRS_ST_PHI))
{
- v[0] = sqrt (Xp / pt->total);
- v[1] = sqrt (Xp / (pt->total * (q - 1)));
+ v[0] = sqrt (Xp / xt->total);
+ v[1] = sqrt (Xp / (xt->total * (q - 1)));
}
if (proc->statistics & (1u << CRS_ST_CC))
- v[2] = sqrt (Xp / (Xp + pt->total));
+ v[2] = sqrt (Xp / (Xp + xt->total));
}
if (proc->statistics & ((1u << CRS_ST_BTAU) | (1u << CRS_ST_CTAU)
double btau_var;
int r, c;
- Dr = Dc = pow2 (pt->total);
- for (r = 0; r < pt->n_rows; r++)
- Dr -= pow2 (pt->row_tot[r]);
- for (c = 0; c < pt->n_cols; c++)
- Dc -= pow2 (pt->col_tot[c]);
+ Dr = Dc = pow2 (xt->total);
+ for (r = 0; r < n_rows; r++)
+ Dr -= pow2 (xt->row_tot[r]);
+ for (c = 0; c < n_cols; c++)
+ Dc -= pow2 (xt->col_tot[c]);
- cum = xnmalloc (pt->n_cols * pt->n_rows, sizeof *cum);
- for (c = 0; c < pt->n_cols; c++)
+ cum = xnmalloc (n_cols * n_rows, sizeof *cum);
+ for (c = 0; c < n_cols; c++)
{
double ct = 0.;
- for (r = 0; r < pt->n_rows; r++)
- cum[c + r * pt->n_cols] = ct += pt->mat[c + r * pt->n_cols];
+ for (r = 0; r < n_rows; r++)
+ cum[c + r * n_cols] = ct += xt->mat[c + r * n_cols];
}
/* P and Q. */
double Cij, Dij;
P = Q = 0.;
- for (i = 0; i < pt->n_rows; i++)
+ for (i = 0; i < n_rows; i++)
{
Cij = Dij = 0.;
- for (j = 1; j < pt->n_cols; j++)
- Cij += pt->col_tot[j] - cum[j + i * pt->n_cols];
+ for (j = 1; j < n_cols; j++)
+ Cij += xt->col_tot[j] - cum[j + i * n_cols];
if (i > 0)
- for (j = 1; j < pt->n_cols; j++)
- Dij += cum[j + (i - 1) * pt->n_cols];
+ for (j = 1; j < n_cols; j++)
+ Dij += cum[j + (i - 1) * n_cols];
for (j = 0;;)
{
- double fij = pt->mat[j + i * pt->n_cols];
+ double fij = xt->mat[j + i * n_cols];
P += fij * Cij;
Q += fij * Dij;
- if (++j == pt->n_cols)
+ if (++j == n_cols)
break;
- assert (j < pt->n_cols);
+ assert (j < n_cols);
- Cij -= pt->col_tot[j] - cum[j + i * pt->n_cols];
- Dij += pt->col_tot[j - 1] - cum[j - 1 + i * pt->n_cols];
+ Cij -= xt->col_tot[j] - cum[j + i * n_cols];
+ Dij += xt->col_tot[j - 1] - cum[j - 1 + i * n_cols];
if (i > 0)
{
- Cij += cum[j - 1 + (i - 1) * pt->n_cols];
- Dij -= cum[j + (i - 1) * pt->n_cols];
+ Cij += cum[j - 1 + (i - 1) * n_cols];
+ Dij -= cum[j + (i - 1) * n_cols];
}
}
}
if (proc->statistics & (1u << CRS_ST_BTAU))
v[3] = (P - Q) / sqrt (Dr * Dc);
if (proc->statistics & (1u << CRS_ST_CTAU))
- v[4] = (q * (P - Q)) / (pow2 (pt->total) * (q - 1));
+ v[4] = (q * (P - Q)) / (pow2 (xt->total) * (q - 1));
if (proc->statistics & (1u << CRS_ST_GAMMA))
v[5] = (P - Q) / (P + Q);
double Cij, Dij;
btau_cum = ctau_cum = gamma_cum = d_yx_cum = d_xy_cum = 0.;
- for (i = 0; i < pt->n_rows; i++)
+ for (i = 0; i < n_rows; i++)
{
Cij = Dij = 0.;
- for (j = 1; j < pt->n_cols; j++)
- Cij += pt->col_tot[j] - cum[j + i * pt->n_cols];
+ for (j = 1; j < n_cols; j++)
+ Cij += xt->col_tot[j] - cum[j + i * n_cols];
if (i > 0)
- for (j = 1; j < pt->n_cols; j++)
- Dij += cum[j + (i - 1) * pt->n_cols];
+ for (j = 1; j < n_cols; j++)
+ Dij += cum[j + (i - 1) * n_cols];
for (j = 0;;)
{
- double fij = pt->mat[j + i * pt->n_cols];
+ double fij = xt->mat[j + i * n_cols];
if (proc->statistics & (1u << CRS_ST_BTAU))
{
const double temp = (2. * sqrt (Dr * Dc) * (Cij - Dij)
- + v[3] * (pt->row_tot[i] * Dc
- + pt->col_tot[j] * Dr));
+ + v[3] * (xt->row_tot[i] * Dc
+ + xt->col_tot[j] * Dr));
btau_cum += fij * temp * temp;
}
if (proc->statistics & (1u << CRS_ST_D))
{
d_yx_cum += fij * pow2 (Dr * (Cij - Dij)
- - (P - Q) * (pt->total - pt->row_tot[i]));
+ - (P - Q) * (xt->total - xt->row_tot[i]));
d_xy_cum += fij * pow2 (Dc * (Dij - Cij)
- - (Q - P) * (pt->total - pt->col_tot[j]));
+ - (Q - P) * (xt->total - xt->col_tot[j]));
}
- if (++j == pt->n_cols)
+ if (++j == n_cols)
break;
- assert (j < pt->n_cols);
+ assert (j < n_cols);
- Cij -= pt->col_tot[j] - cum[j + i * pt->n_cols];
- Dij += pt->col_tot[j - 1] - cum[j - 1 + i * pt->n_cols];
+ Cij -= xt->col_tot[j] - cum[j + i * n_cols];
+ Dij += xt->col_tot[j - 1] - cum[j - 1 + i * n_cols];
if (i > 0)
{
- Cij += cum[j - 1 + (i - 1) * pt->n_cols];
- Dij -= cum[j + (i - 1) * pt->n_cols];
+ Cij += cum[j - 1 + (i - 1) * n_cols];
+ Dij -= cum[j + (i - 1) * n_cols];
}
}
}
}
btau_var = ((btau_cum
- - (pt->total * pow2 (pt->total * (P - Q) / sqrt (Dr * Dc) * (Dr + Dc))))
+ - (xt->total * pow2 (xt->total * (P - Q) / sqrt (Dr * Dc) * (Dr + Dc))))
/ pow2 (Dr * Dc));
if (proc->statistics & (1u << CRS_ST_BTAU))
{
ase[3] = sqrt (btau_var);
- t[3] = v[3] / (2 * sqrt ((ctau_cum - (P - Q) * (P - Q) / pt->total)
+ t[3] = v[3] / (2 * sqrt ((ctau_cum - (P - Q) * (P - Q) / xt->total)
/ (Dr * Dc)));
}
if (proc->statistics & (1u << CRS_ST_CTAU))
{
- ase[4] = ((2 * q / ((q - 1) * pow2 (pt->total)))
- * sqrt (ctau_cum - (P - Q) * (P - Q) / pt->total));
+ ase[4] = ((2 * q / ((q - 1) * pow2 (xt->total)))
+ * sqrt (ctau_cum - (P - Q) * (P - Q) / xt->total));
t[4] = v[4] / ase[4];
}
if (proc->statistics & (1u << CRS_ST_GAMMA))
{
ase[5] = ((4. / ((P + Q) * (P + Q))) * sqrt (gamma_cum));
t[5] = v[5] / (2. / (P + Q)
- * sqrt (ctau_cum - (P - Q) * (P - Q) / pt->total));
+ * sqrt (ctau_cum - (P - Q) * (P - Q) / xt->total));
}
if (proc->statistics & (1u << CRS_ST_D))
{
somers_d_v[0] = (P - Q) / (.5 * (Dc + Dr));
- somers_d_ase[0] = 2. * btau_var / (Dr + Dc) * sqrt (Dr * Dc);
+ somers_d_ase[0] = SYSMIS;
somers_d_t[0] = (somers_d_v[0]
/ (4 / (Dc + Dr)
- * sqrt (ctau_cum - pow2 (P - Q) / pt->total)));
+ * sqrt (ctau_cum - pow2 (P - Q) / xt->total)));
somers_d_v[1] = (P - Q) / Dc;
somers_d_ase[1] = 2. / pow2 (Dc) * sqrt (d_xy_cum);
somers_d_t[1] = (somers_d_v[1]
/ (2. / Dc
- * sqrt (ctau_cum - pow2 (P - Q) / pt->total)));
+ * sqrt (ctau_cum - pow2 (P - Q) / xt->total)));
somers_d_v[2] = (P - Q) / Dr;
somers_d_ase[2] = 2. / pow2 (Dr) * sqrt (d_yx_cum);
somers_d_t[2] = (somers_d_v[2]
/ (2. / Dr
- * sqrt (ctau_cum - pow2 (P - Q) / pt->total)));
+ * sqrt (ctau_cum - pow2 (P - Q) / xt->total)));
}
free (cum);
/* Spearman correlation, Pearson's r. */
if (proc->statistics & (1u << CRS_ST_CORR))
{
- double *R = xmalloc (sizeof *R * pt->n_rows);
- double *C = xmalloc (sizeof *C * pt->n_cols);
+ double *R = xmalloc (sizeof *R * n_rows);
+ double *C = xmalloc (sizeof *C * n_cols);
{
double y, t, c = 0., s = 0.;
for (;;)
{
- R[i] = s + (pt->row_tot[i] + 1.) / 2.;
- y = pt->row_tot[i] - c;
+ R[i] = s + (xt->row_tot[i] + 1.) / 2.;
+ y = xt->row_tot[i] - c;
t = s + y;
c = (t - s) - y;
s = t;
- if (++i == pt->n_rows)
+ if (++i == n_rows)
break;
- assert (i < pt->n_rows);
+ assert (i < n_rows);
}
}
for (;;)
{
- C[j] = s + (pt->col_tot[j] + 1.) / 2;
- y = pt->col_tot[j] - c;
+ C[j] = s + (xt->col_tot[j] + 1.) / 2;
+ y = xt->col_tot[j] - c;
t = s + y;
c = (t - s) - y;
s = t;
- if (++j == pt->n_cols)
+ if (++j == n_cols)
break;
- assert (j < pt->n_cols);
+ assert (j < n_cols);
}
}
- calc_r (pt, R, C, &v[6], &t[6], &ase[6]);
- t[6] = v[6] / t[6];
+ calc_r (xt, R, C, &v[6], &t[6], &ase[6]);
free (R);
free (C);
- calc_r (pt, (double *) pt->rows, (double *) pt->cols, &v[7], &t[7], &ase[7]);
- t[7] = v[7] / t[7];
+ calc_r (xt, (double *) xt->vars[ROW_VAR].values,
+ (double *) xt->vars[COL_VAR].values,
+ &v[7], &t[7], &ase[7]);
}
/* Cohen's kappa. */
- if (proc->statistics & (1u << CRS_ST_KAPPA) && pt->ns_rows == pt->ns_cols)
+ if (proc->statistics & (1u << CRS_ST_KAPPA) && xt->ns_rows == xt->ns_cols)
{
+ double ase_under_h0;
double sum_fii, sum_rici, sum_fiiri_ci, sum_fijri_ci2, sum_riciri_ci;
int i, j;
for (sum_fii = sum_rici = sum_fiiri_ci = sum_riciri_ci = 0., i = j = 0;
- i < pt->ns_rows; i++, j++)
+ i < xt->ns_rows; i++, j++)
{
double prod, sum;
- while (pt->col_tot[j] == 0.)
+ while (xt->col_tot[j] == 0.)
j++;
- prod = pt->row_tot[i] * pt->col_tot[j];
- sum = pt->row_tot[i] + pt->col_tot[j];
+ prod = xt->row_tot[i] * xt->col_tot[j];
+ sum = xt->row_tot[i] + xt->col_tot[j];
- sum_fii += pt->mat[j + i * pt->n_cols];
+ sum_fii += xt->mat[j + i * n_cols];
sum_rici += prod;
- sum_fiiri_ci += pt->mat[j + i * pt->n_cols] * sum;
+ sum_fiiri_ci += xt->mat[j + i * n_cols] * sum;
sum_riciri_ci += prod * sum;
}
- for (sum_fijri_ci2 = 0., i = 0; i < pt->ns_rows; i++)
- for (j = 0; j < pt->ns_cols; j++)
+ for (sum_fijri_ci2 = 0., i = 0; i < xt->ns_rows; i++)
+ for (j = 0; j < xt->ns_cols; j++)
{
- double sum = pt->row_tot[i] + pt->col_tot[j];
- sum_fijri_ci2 += pt->mat[j + i * pt->n_cols] * sum * sum;
+ double sum = xt->row_tot[i] + xt->col_tot[j];
+ sum_fijri_ci2 += xt->mat[j + i * n_cols] * sum * sum;
}
- v[8] = (pt->total * sum_fii - sum_rici) / (pow2 (pt->total) - sum_rici);
+ v[8] = (xt->total * sum_fii - sum_rici) / (pow2 (xt->total) - sum_rici);
+
+ ase_under_h0 = sqrt ((pow2 (xt->total) * sum_rici
+ + sum_rici * sum_rici
+ - xt->total * sum_riciri_ci)
+ / (xt->total * (pow2 (xt->total) - sum_rici) * (pow2 (xt->total) - sum_rici)));
- ase[8] = sqrt ((pow2 (pt->total) * sum_rici
- + sum_rici * sum_rici
- - pt->total * sum_riciri_ci)
- / (pt->total * (pow2 (pt->total) - sum_rici) * (pow2 (pt->total) - sum_rici)));
-#if 0
- t[8] = v[8] / sqrt (pt->total * (((sum_fii * (pt->total - sum_fii))
- / pow2 (pow2 (pt->total) - sum_rici))
- + ((2. * (pt->total - sum_fii)
+ ase[8] = sqrt (xt->total * (((sum_fii * (xt->total - sum_fii))
+ / pow2 (pow2 (xt->total) - sum_rici))
+ + ((2. * (xt->total - sum_fii)
* (2. * sum_fii * sum_rici
- - pt->total * sum_fiiri_ci))
- / cube (pow2 (pt->total) - sum_rici))
- + (pow2 (pt->total - sum_fii)
- * (pt->total * sum_fijri_ci2 - 4.
+ - xt->total * sum_fiiri_ci))
+ / pow3 (pow2 (xt->total) - sum_rici))
+ + (pow2 (xt->total - sum_fii)
+ * (xt->total * sum_fijri_ci2 - 4.
* sum_rici * sum_rici)
- / pow4 (pow2 (pt->total) - sum_rici))));
-#else
- t[8] = v[8] / ase[8];
-#endif
+ / pow4 (pow2 (xt->total) - sum_rici))));
+
+ t[8] = v[8] / ase_under_h0;
}
return 1;
/* Calculate risk estimate. */
static int
-calc_risk (struct pivot_table *pt,
+calc_risk (struct crosstabulation *xt,
double *value, double *upper, double *lower, union value *c)
{
+ size_t n_cols = xt->vars[COL_VAR].n_values;
double f11, f12, f21, f22;
double v;
value[i] = upper[i] = lower[i] = SYSMIS;
}
- if (pt->ns_rows != 2 || pt->ns_cols != 2)
+ if (xt->ns_rows != 2 || xt->ns_cols != 2)
return 0;
{
int nz_cols[2];
int i, j;
- for (i = j = 0; i < pt->n_cols; i++)
- if (pt->col_tot[i] != 0.)
+ for (i = j = 0; i < n_cols; i++)
+ if (xt->col_tot[i] != 0.)
{
nz_cols[j++] = i;
if (j == 2)
assert (j == 2);
- f11 = pt->mat[nz_cols[0]];
- f12 = pt->mat[nz_cols[1]];
- f21 = pt->mat[nz_cols[0] + pt->n_cols];
- f22 = pt->mat[nz_cols[1] + pt->n_cols];
+ f11 = xt->mat[nz_cols[0]];
+ f12 = xt->mat[nz_cols[1]];
+ f21 = xt->mat[nz_cols[0] + n_cols];
+ f22 = xt->mat[nz_cols[1] + n_cols];
- c[0] = pt->cols[nz_cols[0]];
- c[1] = pt->cols[nz_cols[1]];
+ c[0] = xt->vars[COL_VAR].values[nz_cols[0]];
+ c[1] = xt->vars[COL_VAR].values[nz_cols[1]];
}
value[0] = (f11 * f22) / (f12 * f21);
/* Calculate directional measures. */
static int
-calc_directional (struct crosstabs_proc *proc, struct pivot_table *pt,
+calc_directional (struct crosstabs_proc *proc, struct crosstabulation *xt,
double v[N_DIRECTIONAL], double ase[N_DIRECTIONAL],
- double t[N_DIRECTIONAL])
+ double t[N_DIRECTIONAL], double sig[N_DIRECTIONAL])
{
+ size_t n_rows = xt->vars[ROW_VAR].n_values;
+ size_t n_cols = xt->vars[COL_VAR].n_values;
{
int i;
for (i = 0; i < N_DIRECTIONAL; i++)
- v[i] = ase[i] = t[i] = SYSMIS;
+ v[i] = ase[i] = t[i] = sig[i] = SYSMIS;
}
/* Lambda. */
if (proc->statistics & (1u << CRS_ST_LAMBDA))
{
- double *fim = xnmalloc (pt->n_rows, sizeof *fim);
- int *fim_index = xnmalloc (pt->n_rows, sizeof *fim_index);
- double *fmj = xnmalloc (pt->n_cols, sizeof *fmj);
- int *fmj_index = xnmalloc (pt->n_cols, sizeof *fmj_index);
+ double *fim = xnmalloc (n_rows, sizeof *fim);
+ int *fim_index = xnmalloc (n_rows, sizeof *fim_index);
+ double *fmj = xnmalloc (n_cols, sizeof *fmj);
+ int *fmj_index = xnmalloc (n_cols, sizeof *fmj_index);
double sum_fim, sum_fmj;
double rm, cm;
int rm_index, cm_index;
int i, j;
/* Find maximum for each row and their sum. */
- for (sum_fim = 0., i = 0; i < pt->n_rows; i++)
+ for (sum_fim = 0., i = 0; i < n_rows; i++)
{
- double max = pt->mat[i * pt->n_cols];
+ double max = xt->mat[i * n_cols];
int index = 0;
- for (j = 1; j < pt->n_cols; j++)
- if (pt->mat[j + i * pt->n_cols] > max)
+ for (j = 1; j < n_cols; j++)
+ if (xt->mat[j + i * n_cols] > max)
{
- max = pt->mat[j + i * pt->n_cols];
+ max = xt->mat[j + i * n_cols];
index = j;
}
}
/* Find maximum for each column. */
- for (sum_fmj = 0., j = 0; j < pt->n_cols; j++)
+ for (sum_fmj = 0., j = 0; j < n_cols; j++)
{
- double max = pt->mat[j];
+ double max = xt->mat[j];
int index = 0;
- for (i = 1; i < pt->n_rows; i++)
- if (pt->mat[j + i * pt->n_cols] > max)
+ for (i = 1; i < n_rows; i++)
+ if (xt->mat[j + i * n_cols] > max)
{
- max = pt->mat[j + i * pt->n_cols];
+ max = xt->mat[j + i * n_cols];
index = i;
}
}
/* Find maximum row total. */
- rm = pt->row_tot[0];
+ rm = xt->row_tot[0];
rm_index = 0;
- for (i = 1; i < pt->n_rows; i++)
- if (pt->row_tot[i] > rm)
+ for (i = 1; i < n_rows; i++)
+ if (xt->row_tot[i] > rm)
{
- rm = pt->row_tot[i];
+ rm = xt->row_tot[i];
rm_index = i;
}
/* Find maximum column total. */
- cm = pt->col_tot[0];
+ cm = xt->col_tot[0];
cm_index = 0;
- for (j = 1; j < pt->n_cols; j++)
- if (pt->col_tot[j] > cm)
+ for (j = 1; j < n_cols; j++)
+ if (xt->col_tot[j] > cm)
{
- cm = pt->col_tot[j];
+ cm = xt->col_tot[j];
cm_index = j;
}
- v[0] = (sum_fim + sum_fmj - cm - rm) / (2. * pt->total - rm - cm);
- v[1] = (sum_fmj - rm) / (pt->total - rm);
- v[2] = (sum_fim - cm) / (pt->total - cm);
+ v[0] = (sum_fim + sum_fmj - cm - rm) / (2. * xt->total - rm - cm);
+ v[1] = (sum_fmj - rm) / (xt->total - rm);
+ v[2] = (sum_fim - cm) / (xt->total - cm);
- /* ASE1 for Y given PT. */
+ /* ASE1 for Y given XT. */
{
- double accum;
-
- for (accum = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
- {
- const int deltaj = j == cm_index;
- accum += (pt->mat[j + i * pt->n_cols]
- * pow2 ((j == fim_index[i])
- - deltaj
- + v[0] * deltaj));
- }
-
- ase[2] = sqrt (accum - pt->total * v[0]) / (pt->total - cm);
+ double accum;
+
+ accum = 0.;
+ for (i = 0; i < n_rows; i++)
+ if (cm_index == fim_index[i])
+ accum += fim[i];
+ ase[2] = sqrt ((xt->total - sum_fim) * (sum_fim + cm - 2. * accum)
+ / pow3 (xt->total - cm));
}
- /* ASE0 for Y given PT. */
+ /* ASE0 for Y given XT. */
{
double accum;
- for (accum = 0., i = 0; i < pt->n_rows; i++)
+ for (accum = 0., i = 0; i < n_rows; i++)
if (cm_index != fim_index[i])
- accum += (pt->mat[i * pt->n_cols + fim_index[i]]
- + pt->mat[i * pt->n_cols + cm_index]);
- t[2] = v[2] / (sqrt (accum - pow2 (sum_fim - cm) / pt->total) / (pt->total - cm));
+ accum += (xt->mat[i * n_cols + fim_index[i]]
+ + xt->mat[i * n_cols + cm_index]);
+ t[2] = v[2] / (sqrt (accum - pow2 (sum_fim - cm) / xt->total) / (xt->total - cm));
}
- /* ASE1 for PT given Y. */
+ /* ASE1 for XT given Y. */
{
- double accum;
-
- for (accum = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
- {
- const int deltaj = i == rm_index;
- accum += (pt->mat[j + i * pt->n_cols]
- * pow2 ((i == fmj_index[j])
- - deltaj
- + v[0] * deltaj));
- }
-
- ase[1] = sqrt (accum - pt->total * v[0]) / (pt->total - rm);
+ double accum;
+
+ accum = 0.;
+ for (j = 0; j < n_cols; j++)
+ if (rm_index == fmj_index[j])
+ accum += fmj[j];
+ ase[1] = sqrt ((xt->total - sum_fmj) * (sum_fmj + rm - 2. * accum)
+ / pow3 (xt->total - rm));
}
- /* ASE0 for PT given Y. */
+ /* ASE0 for XT given Y. */
{
double accum;
- for (accum = 0., j = 0; j < pt->n_cols; j++)
+ for (accum = 0., j = 0; j < n_cols; j++)
if (rm_index != fmj_index[j])
- accum += (pt->mat[j + pt->n_cols * fmj_index[j]]
- + pt->mat[j + pt->n_cols * rm_index]);
- t[1] = v[1] / (sqrt (accum - pow2 (sum_fmj - rm) / pt->total) / (pt->total - rm));
+ accum += (xt->mat[j + n_cols * fmj_index[j]]
+ + xt->mat[j + n_cols * rm_index]);
+ t[1] = v[1] / (sqrt (accum - pow2 (sum_fmj - rm) / xt->total) / (xt->total - rm));
}
/* Symmetric ASE0 and ASE1. */
double accum0;
double accum1;
- for (accum0 = accum1 = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
+ for (accum0 = accum1 = 0., i = 0; i < n_rows; i++)
+ for (j = 0; j < n_cols; j++)
{
int temp0 = (fmj_index[j] == i) + (fim_index[i] == j);
int temp1 = (i == rm_index) + (j == cm_index);
- accum0 += pt->mat[j + i * pt->n_cols] * pow2 (temp0 - temp1);
- accum1 += (pt->mat[j + i * pt->n_cols]
+ accum0 += xt->mat[j + i * n_cols] * pow2 (temp0 - temp1);
+ accum1 += (xt->mat[j + i * n_cols]
* pow2 (temp0 + (v[0] - 1.) * temp1));
}
- ase[0] = sqrt (accum1 - 4. * pt->total * v[0] * v[0]) / (2. * pt->total - rm - cm);
- t[0] = v[0] / (sqrt (accum0 - pow2 ((sum_fim + sum_fmj - cm - rm) / pt->total))
- / (2. * pt->total - rm - cm));
+ ase[0] = sqrt (accum1 - 4. * xt->total * v[0] * v[0]) / (2. * xt->total - rm - cm);
+ t[0] = v[0] / (sqrt (accum0 - pow2 (sum_fim + sum_fmj - cm - rm) / xt->total)
+ / (2. * xt->total - rm - cm));
}
+ for (i = 0; i < 3; i++)
+ sig[i] = 2 * gsl_cdf_ugaussian_Q (t[i]);
+
free (fim);
free (fim_index);
free (fmj);
free (fmj_index);
+ /* Tau. */
{
double sum_fij2_ri, sum_fij2_ci;
double sum_ri2, sum_cj2;
- for (sum_fij2_ri = sum_fij2_ci = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
+ for (sum_fij2_ri = sum_fij2_ci = 0., i = 0; i < n_rows; i++)
+ for (j = 0; j < n_cols; j++)
{
- double temp = pow2 (pt->mat[j + i * pt->n_cols]);
- sum_fij2_ri += temp / pt->row_tot[i];
- sum_fij2_ci += temp / pt->col_tot[j];
+ double temp = pow2 (xt->mat[j + i * n_cols]);
+ sum_fij2_ri += temp / xt->row_tot[i];
+ sum_fij2_ci += temp / xt->col_tot[j];
}
- for (sum_ri2 = 0., i = 0; i < pt->n_rows; i++)
- sum_ri2 += pow2 (pt->row_tot[i]);
+ for (sum_ri2 = 0., i = 0; i < n_rows; i++)
+ sum_ri2 += pow2 (xt->row_tot[i]);
- for (sum_cj2 = 0., j = 0; j < pt->n_cols; j++)
- sum_cj2 += pow2 (pt->col_tot[j]);
+ for (sum_cj2 = 0., j = 0; j < n_cols; j++)
+ sum_cj2 += pow2 (xt->col_tot[j]);
- v[3] = (pt->total * sum_fij2_ci - sum_ri2) / (pow2 (pt->total) - sum_ri2);
- v[4] = (pt->total * sum_fij2_ri - sum_cj2) / (pow2 (pt->total) - sum_cj2);
+ v[3] = (xt->total * sum_fij2_ci - sum_ri2) / (pow2 (xt->total) - sum_ri2);
+ v[4] = (xt->total * sum_fij2_ri - sum_cj2) / (pow2 (xt->total) - sum_cj2);
}
}
double ase1_yx, ase1_xy, ase1_sym;
int i, j;
- for (UX = 0., i = 0; i < pt->n_rows; i++)
- if (pt->row_tot[i] > 0.)
- UX -= pt->row_tot[i] / pt->total * log (pt->row_tot[i] / pt->total);
+ for (UX = 0., i = 0; i < n_rows; i++)
+ if (xt->row_tot[i] > 0.)
+ UX -= xt->row_tot[i] / xt->total * log (xt->row_tot[i] / xt->total);
- for (UY = 0., j = 0; j < pt->n_cols; j++)
- if (pt->col_tot[j] > 0.)
- UY -= pt->col_tot[j] / pt->total * log (pt->col_tot[j] / pt->total);
+ for (UY = 0., j = 0; j < n_cols; j++)
+ if (xt->col_tot[j] > 0.)
+ UY -= xt->col_tot[j] / xt->total * log (xt->col_tot[j] / xt->total);
- for (UXY = P = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
+ for (UXY = P = 0., i = 0; i < n_rows; i++)
+ for (j = 0; j < n_cols; j++)
{
- double entry = pt->mat[j + i * pt->n_cols];
+ double entry = xt->mat[j + i * n_cols];
if (entry <= 0.)
continue;
- P += entry * pow2 (log (pt->col_tot[j] * pt->row_tot[i] / (pt->total * entry)));
- UXY -= entry / pt->total * log (entry / pt->total);
+ P += entry * pow2 (log (xt->col_tot[j] * xt->row_tot[i] / (xt->total * entry)));
+ UXY -= entry / xt->total * log (entry / xt->total);
}
- for (ase1_yx = ase1_xy = ase1_sym = 0., i = 0; i < pt->n_rows; i++)
- for (j = 0; j < pt->n_cols; j++)
+ for (ase1_yx = ase1_xy = ase1_sym = 0., i = 0; i < n_rows; i++)
+ for (j = 0; j < n_cols; j++)
{
- double entry = pt->mat[j + i * pt->n_cols];
+ double entry = xt->mat[j + i * n_cols];
if (entry <= 0.)
continue;
- ase1_yx += entry * pow2 (UY * log (entry / pt->row_tot[i])
- + (UX - UXY) * log (pt->col_tot[j] / pt->total));
- ase1_xy += entry * pow2 (UX * log (entry / pt->col_tot[j])
- + (UY - UXY) * log (pt->row_tot[i] / pt->total));
+ ase1_yx += entry * pow2 (UY * log (entry / xt->row_tot[i])
+ + (UX - UXY) * log (xt->col_tot[j] / xt->total));
+ ase1_xy += entry * pow2 (UX * log (entry / xt->col_tot[j])
+ + (UY - UXY) * log (xt->row_tot[i] / xt->total));
ase1_sym += entry * pow2 ((UXY
- * log (pt->row_tot[i] * pt->col_tot[j] / pow2 (pt->total)))
- - (UX + UY) * log (entry / pt->total));
+ * log (xt->row_tot[i] * xt->col_tot[j] / pow2 (xt->total)))
+ - (UX + UY) * log (entry / xt->total));
}
v[5] = 2. * ((UX + UY - UXY) / (UX + UY));
- ase[5] = (2. / (pt->total * pow2 (UX + UY))) * sqrt (ase1_sym);
- t[5] = v[5] / ((2. / (pt->total * (UX + UY)))
- * sqrt (P - pow2 (UX + UY - UXY) / pt->total));
+ ase[5] = (2. / (xt->total * pow2 (UX + UY))) * sqrt (ase1_sym);
+ t[5] = SYSMIS;
v[6] = (UX + UY - UXY) / UX;
- ase[6] = sqrt (ase1_xy) / (pt->total * UX * UX);
- t[6] = v[6] / (sqrt (P - pt->total * pow2 (UX + UY - UXY)) / (pt->total * UX));
+ ase[6] = sqrt (ase1_xy) / (xt->total * UX * UX);
+ t[6] = v[6] / (sqrt (P - xt->total * pow2 (UX + UY - UXY)) / (xt->total * UX));
v[7] = (UX + UY - UXY) / UY;
- ase[7] = sqrt (ase1_yx) / (pt->total * UY * UY);
- t[7] = v[7] / (sqrt (P - pt->total * pow2 (UX + UY - UXY)) / (pt->total * UY));
+ ase[7] = sqrt (ase1_yx) / (xt->total * UY * UY);
+ t[7] = v[7] / (sqrt (P - xt->total * pow2 (UX + UY - UXY)) / (xt->total * UY));
}
/* Somers' D. */
double somers_d_ase[3];
double somers_d_t[3];
- if (calc_symmetric (proc, pt, v_dummy, ase_dummy, t_dummy,
+ if (calc_symmetric (proc, xt, v_dummy, ase_dummy, t_dummy,
somers_d_v, somers_d_ase, somers_d_t))
{
int i;
v[8 + i] = somers_d_v[i];
ase[8 + i] = somers_d_ase[i];
t[8 + i] = somers_d_t[i];
+ sig[8 + i] = 2 * gsl_cdf_ugaussian_Q (fabs (somers_d_t[i]));
}
}
}
double SX, SXW;
int i, j;
- for (sum_Xr = sum_X2r = 0., i = 0; i < pt->n_rows; i++)
+ for (sum_Xr = sum_X2r = 0., i = 0; i < n_rows; i++)
{
- sum_Xr += pt->rows[i].f * pt->row_tot[i];
- sum_X2r += pow2 (pt->rows[i].f) * pt->row_tot[i];
+ sum_Xr += xt->vars[ROW_VAR].values[i].f * xt->row_tot[i];
+ sum_X2r += pow2 (xt->vars[ROW_VAR].values[i].f) * xt->row_tot[i];
}
- SX = sum_X2r - pow2 (sum_Xr) / pt->total;
+ SX = sum_X2r - pow2 (sum_Xr) / xt->total;
- for (SXW = 0., j = 0; j < pt->n_cols; j++)
+ for (SXW = 0., j = 0; j < n_cols; j++)
{
double cum;
- for (cum = 0., i = 0; i < pt->n_rows; i++)
+ for (cum = 0., i = 0; i < n_rows; i++)
{
- SXW += pow2 (pt->rows[i].f) * pt->mat[j + i * pt->n_cols];
- cum += pt->rows[i].f * pt->mat[j + i * pt->n_cols];
+ SXW += (pow2 (xt->vars[ROW_VAR].values[i].f)
+ * xt->mat[j + i * n_cols]);
+ cum += (xt->vars[ROW_VAR].values[i].f
+ * xt->mat[j + i * n_cols]);
}
- SXW -= cum * cum / pt->col_tot[j];
+ SXW -= cum * cum / xt->col_tot[j];
}
v[11] = sqrt (1. - SXW / SX);
}
double SY, SYW;
int i, j;
- for (sum_Yc = sum_Y2c = 0., i = 0; i < pt->n_cols; i++)
+ for (sum_Yc = sum_Y2c = 0., i = 0; i < n_cols; i++)
{
- sum_Yc += pt->cols[i].f * pt->col_tot[i];
- sum_Y2c += pow2 (pt->cols[i].f) * pt->col_tot[i];
+ sum_Yc += xt->vars[COL_VAR].values[i].f * xt->col_tot[i];
+ sum_Y2c += pow2 (xt->vars[COL_VAR].values[i].f) * xt->col_tot[i];
}
- SY = sum_Y2c - sum_Yc * sum_Yc / pt->total;
+ SY = sum_Y2c - sum_Yc * sum_Yc / xt->total;
- for (SYW = 0., i = 0; i < pt->n_rows; i++)
+ for (SYW = 0., i = 0; i < n_rows; i++)
{
double cum;
- for (cum = 0., j = 0; j < pt->n_cols; j++)
+ for (cum = 0., j = 0; j < n_cols; j++)
{
- SYW += pow2 (pt->cols[j].f) * pt->mat[j + i * pt->n_cols];
- cum += pt->cols[j].f * pt->mat[j + i * pt->n_cols];
+ SYW += (pow2 (xt->vars[COL_VAR].values[j].f)
+ * xt->mat[j + i * n_cols]);
+ cum += (xt->vars[COL_VAR].values[j].f
+ * xt->mat[j + i * n_cols]);
}
- SYW -= cum * cum / pt->row_tot[i];
+ SYW -= cum * cum / xt->row_tot[i];
}
v[12] = sqrt (1. - SYW / SY);
}
projects / pspp / blobdiff