/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2007, 2009, 2010 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2007, 2009, 2010, 2011 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
#include "data/case.h"
#include "data/casegrouper.h"
#include "data/casereader.h"
+#include "data/dataset.h"
#include "data/dictionary.h"
#include "data/format.h"
-#include "data/procedure.h"
#include "data/settings.h"
#include "data/value-labels.h"
#include "data/variable.h"
{
double p; /* the %ile to be calculated */
double value; /* the %ile's value */
- double x1; /* The datum value <= the percentile */
- double x2; /* The datum value >= the percentile */
- int flag;
- int flag2; /* Set to 1 if this percentile value has been found */
bool show; /* True to show this percentile in the statistics box. */
};
-
-static struct percentile *percentiles;
-static int n_percentiles, n_show_percentiles;
-
/* Groups of statistics. */
#define BI BIT_INDEX
#define FRQ_DEFAULT \
| BI(FRQ_SKEW) | BI(FRQ_SESKEW) | BI(FRQ_RANGE) \
| BI(FRQ_RANGE) | BI(FRQ_MODE) | BI(FRQ_MEDIAN))
-/* Statistics; number of statistics. */
-static unsigned long stats;
-static int n_stats;
-
struct frq_chart
{
double x_min; /* X axis minimum value. */
bool include_missing; /* Whether to include missing values. */
};
-/* Histogram and pie chart settings. */
-static struct frq_chart hist, pie;
-
-/* Parsed command. */
-static struct cmd_frequencies cmd;
-
/* Frequency tables. */
/* Entire frequency table. */
struct var_freqs *vars;
size_t n_vars;
+
+ /* Percentiles to calculate and possibly display. */
+ struct percentile *percentiles;
+ int n_percentiles, n_show_percentiles;
+
+ /* Frequency table display. */
+ int max_categories; /* Maximum categories to show. */
+ int sort; /* FRQ_AVALUE or FRQ_DVALUE
+ or FRQ_ACOUNT or FRQ_DCOUNT. */
+
+ /* Statistics; number of statistics. */
+ unsigned long stats;
+ int n_stats;
+
+ /* Histogram and pie chart settings. */
+ struct frq_chart *hist, *pie;
};
-static void determine_charts (void);
+static void determine_charts (struct frq_proc *,
+ const struct cmd_frequencies *);
-static void calc_stats (const struct var_freqs *v, double d[FRQ_N_STATS]);
+static void calc_stats (const struct var_freqs *, double d[FRQ_N_STATS]);
+static void calc_percentiles (const struct frq_proc *,
+ const struct var_freqs *);
static void precalc (struct frq_proc *, struct casereader *, struct dataset *);
static void calc (struct frq_proc *, const struct ccase *,
const struct dataset *);
static void postcalc (struct frq_proc *, const struct dataset *);
-static void postprocess_freq_tab (struct var_freqs *);
+static void postprocess_freq_tab (const struct frq_proc *, struct var_freqs *);
static void dump_freq_table (const struct var_freqs *,
const struct variable *weight_var);
-static void dump_statistics (const struct var_freqs *,
+static void dump_statistics (const struct frq_proc *, const struct var_freqs *,
const struct variable *weight_var);
static void cleanup_freq_tab (struct var_freqs *);
-static algo_compare_func compare_value_numeric_a, compare_value_alpha_a;
-static algo_compare_func compare_value_numeric_d, compare_value_alpha_d;
-static algo_compare_func compare_freq_numeric_a, compare_freq_alpha_a;
-static algo_compare_func compare_freq_numeric_d, compare_freq_alpha_d;
-
-static void add_percentile (struct frq_proc *, double x, bool show);
-
-static void do_piechart(const struct variable *var,
- const struct freq_tab *frq_tab);
-
-struct histogram *
-freq_tab_to_hist(const struct freq_tab *ft, const struct variable *var);
+static void add_percentile (struct frq_proc *, double x, bool show,
+ size_t *allocated_percentiles);
+static void do_piechart(const struct frq_chart *, const struct variable *,
+ const struct freq_tab *);
+struct histogram *freq_tab_to_hist(const struct frq_proc *,
+ const struct freq_tab *,
+ const struct variable *);
\f
/* Parser and outline. */
int
cmd_frequencies (struct lexer *lexer, struct dataset *ds)
{
+ struct cmd_frequencies cmd;
struct frq_proc frq;
struct casegrouper *grouper;
struct casereader *input, *group;
+ size_t allocated_percentiles;
bool ok;
int i;
- n_percentiles = 0;
- n_show_percentiles = 0;
- percentiles = NULL;
-
frq.pool = pool_create ();
+
frq.vars = NULL;
frq.n_vars = 0;
+ frq.percentiles = NULL;
+ frq.n_percentiles = 0;
+ frq.n_show_percentiles = 0;
+
+ frq.hist = NULL;
+ frq.pie = NULL;
+
+ allocated_percentiles = 0;
+
if (!parse_frequencies (lexer, ds, &cmd, &frq))
{
pool_destroy (frq.pool);
return CMD_FAILURE;
}
+ /* Figure out when to show frequency tables. */
+ frq.max_categories = (cmd.table == FRQ_NOTABLE ? -1
+ : cmd.table == FRQ_TABLE ? INT_MAX
+ : cmd.limit);
+ frq.sort = cmd.sort;
+
/* Figure out statistics to calculate. */
- stats = 0;
+ frq.stats = 0;
if (cmd.a_statistics[FRQ_ST_DEFAULT] || !cmd.sbc_statistics)
- stats |= FRQ_DEFAULT;
+ frq.stats |= FRQ_DEFAULT;
if (cmd.a_statistics[FRQ_ST_ALL])
- stats |= FRQ_ALL;
+ frq.stats |= FRQ_ALL;
if (cmd.sort != FRQ_AVALUE && cmd.sort != FRQ_DVALUE)
- stats &= ~BIT_INDEX (FRQ_MEDIAN);
+ frq.stats &= ~BIT_INDEX (FRQ_MEDIAN);
for (i = 0; i < FRQ_N_STATS; i++)
if (cmd.a_statistics[st_name[i].st_indx])
- stats |= BIT_INDEX (i);
- if (stats & FRQ_KURT)
- stats |= BIT_INDEX (FRQ_SEKURT);
- if (stats & FRQ_SKEW)
- stats |= BIT_INDEX (FRQ_SESKEW);
+ frq.stats |= BIT_INDEX (i);
+ if (frq.stats & FRQ_KURT)
+ frq.stats |= BIT_INDEX (FRQ_SEKURT);
+ if (frq.stats & FRQ_SKEW)
+ frq.stats |= BIT_INDEX (FRQ_SESKEW);
/* Calculate n_stats. */
- n_stats = 0;
+ frq.n_stats = 0;
for (i = 0; i < FRQ_N_STATS; i++)
- if ((stats & BIT_INDEX (i)))
- n_stats++;
+ if ((frq.stats & BIT_INDEX (i)))
+ frq.n_stats++;
/* Charting. */
- determine_charts ();
+ determine_charts (&frq, &cmd);
if (cmd.sbc_histogram || cmd.sbc_piechart || cmd.sbc_ntiles)
cmd.sort = FRQ_AVALUE;
subc_list_double *ptl_list = &cmd.dl_percentiles[i];
for ( pl = 0 ; pl < subc_list_double_count(ptl_list); ++pl)
add_percentile (&frq, subc_list_double_at(ptl_list, pl) / 100.0,
- true);
+ true, &allocated_percentiles);
}
}
if ( cmd.sbc_ntiles )
{
int j;
for (j = 0; j <= cmd.n_ntiles[i]; ++j )
- add_percentile (&frq, j / (double) cmd.n_ntiles[i], true);
+ add_percentile (&frq, j / (double) cmd.n_ntiles[i], true,
+ &allocated_percentiles);
}
}
- if (stats & BIT_INDEX (FRQ_MEDIAN))
+ if (frq.stats & BIT_INDEX (FRQ_MEDIAN))
{
/* Treat the median as the 50% percentile.
We output it in the percentiles table as "50 (Median)." */
- add_percentile (&frq, 0.5, true);
- stats &= ~BIT_INDEX (FRQ_MEDIAN);
- n_stats--;
+ add_percentile (&frq, 0.5, true, &allocated_percentiles);
+ frq.stats &= ~BIT_INDEX (FRQ_MEDIAN);
+ frq.n_stats--;
}
if (cmd.sbc_histogram)
{
- add_percentile (&frq, 0.25, false);
- add_percentile (&frq, 0.75, false);
+ add_percentile (&frq, 0.25, false, &allocated_percentiles);
+ add_percentile (&frq, 0.75, false, &allocated_percentiles);
}
/* Do it! */
pool_destroy (frq.pool);
free (frq.vars);
+ free (frq.percentiles);
+ free (frq.hist);
+ free (frq.pie);
return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
}
/* Figure out which charts the user requested. */
static void
-determine_charts (void)
+determine_charts (struct frq_proc *frq, const struct cmd_frequencies *cmd)
{
- if (cmd.sbc_barchart)
+ if (cmd->sbc_barchart)
msg (SW, _("Bar charts are not implemented."));
- if (cmd.sbc_histogram)
+ if (cmd->sbc_histogram)
{
- hist.x_min = cmd.hi_min;
- hist.x_max = cmd.hi_max;
- hist.y_scale = cmd.hi_scale;
- hist.y_max = cmd.hi_scale == FRQ_FREQ ? cmd.hi_freq : cmd.hi_pcnt;
- hist.draw_normal = cmd.hi_norm != FRQ_NONORMAL;
- hist.include_missing = false;
-
- if (hist.x_min != SYSMIS && hist.x_max != SYSMIS
- && hist.x_min >= hist.x_max)
+ struct frq_chart *hist;
+
+ hist = frq->hist = xmalloc (sizeof *frq->hist);
+ hist->x_min = cmd->hi_min;
+ hist->x_max = cmd->hi_max;
+ hist->y_scale = cmd->hi_scale;
+ hist->y_max = cmd->hi_scale == FRQ_FREQ ? cmd->hi_freq : cmd->hi_pcnt;
+ hist->draw_normal = cmd->hi_norm != FRQ_NONORMAL;
+ hist->include_missing = false;
+
+ if (hist->x_min != SYSMIS && hist->x_max != SYSMIS
+ && hist->x_min >= hist->x_max)
{
msg (SE, _("MAX for histogram must be greater than or equal to MIN, "
"but MIN was specified as %.15g and MAX as %.15g. "
- "MIN and MAX will be ignored."), hist.x_min, hist.x_max);
- hist.x_min = hist.x_max = SYSMIS;
+ "MIN and MAX will be ignored."),
+ hist->x_min, hist->x_max);
+ hist->x_min = hist->x_max = SYSMIS;
}
}
- if (cmd.sbc_piechart)
+ if (cmd->sbc_piechart)
{
- pie.x_min = cmd.pie_min;
- pie.x_max = cmd.pie_max;
- pie.y_scale = cmd.pie_scale;
- pie.include_missing = cmd.pie_missing == FRQ_MISSING;
+ struct frq_chart *pie;
- if (pie.x_min != SYSMIS && pie.x_max != SYSMIS
- && pie.x_min >= pie.x_max)
+ pie = frq->pie = xmalloc (sizeof *frq->pie);
+ pie->x_min = cmd->pie_min;
+ pie->x_max = cmd->pie_max;
+ pie->y_scale = cmd->pie_scale;
+ pie->include_missing = cmd->pie_missing == FRQ_MISSING;
+
+ if (pie->x_min != SYSMIS && pie->x_max != SYSMIS
+ && pie->x_min >= pie->x_max)
{
msg (SE, _("MAX for pie chart must be greater than or equal to MIN, "
"but MIN was specified as %.15g and MAX as %.15g. "
- "MIN and MAX will be ignored."), pie.x_min, pie.x_max);
- pie.x_min = pie.x_max = SYSMIS;
+ "MIN and MAX will be ignored."), pie->x_min, pie->x_max);
+ pie->x_min = pie->x_max = SYSMIS;
}
}
-
}
/* Add data from case C to the frequency table. */
for (i = 0; i < frq->n_vars; i++)
{
struct var_freqs *vf = &frq->vars[i];
- int n_categories;
- postprocess_freq_tab (vf);
+ postprocess_freq_tab (frq, vf);
/* Frequencies tables. */
- n_categories = vf->tab.n_valid + vf->tab.n_missing;
- if (cmd.table == FRQ_TABLE
- || (cmd.table == FRQ_LIMIT && n_categories <= cmd.limit))
+ if (vf->tab.n_valid + vf->tab.n_missing <= frq->max_categories)
dump_freq_table (vf, wv);
/* Statistics. */
- if (n_stats)
- dump_statistics (vf, wv);
+ if (frq->n_stats)
+ dump_statistics (frq, vf, wv);
- if (cmd.sbc_histogram && var_is_numeric (vf->var) && vf->tab.n_valid > 0)
+ if (frq->hist && var_is_numeric (vf->var) && vf->tab.n_valid > 0)
{
double d[FRQ_N_STATS];
struct histogram *histogram;
calc_stats (vf, d);
- histogram = freq_tab_to_hist (&vf->tab, vf->var);
+ histogram = freq_tab_to_hist (frq, &vf->tab, vf->var);
chart_item_submit (histogram_chart_create (
histogram->gsl_hist, var_to_string(vf->var),
vf->tab.valid_cases,
d[FRQ_MEAN],
d[FRQ_STDDEV],
- hist.draw_normal));
+ frq->hist->draw_normal));
statistic_destroy (&histogram->parent);
}
- if (cmd.sbc_piechart)
- do_piechart(vf->var, &vf->tab);
+ if (frq->pie)
+ do_piechart(frq->pie, vf->var, &vf->tab);
cleanup_freq_tab (vf);
}
}
-/* Returns the comparison function that should be used for
- sorting a frequency table by FRQ_SORT using VAL_TYPE
- values. */
-static algo_compare_func *
-get_freq_comparator (int frq_sort, enum val_type val_type)
-{
- bool is_numeric = val_type == VAL_NUMERIC;
- switch (frq_sort)
- {
- case FRQ_AVALUE:
- return is_numeric ? compare_value_numeric_a : compare_value_alpha_a;
- case FRQ_DVALUE:
- return is_numeric ? compare_value_numeric_d : compare_value_alpha_d;
- case FRQ_AFREQ:
- return is_numeric ? compare_freq_numeric_a : compare_freq_alpha_a;
- case FRQ_DFREQ:
- return is_numeric ? compare_freq_numeric_d : compare_freq_alpha_d;
- default:
- NOT_REACHED ();
- }
-}
-
/* Returns true iff the value in struct freq F is non-missing
for variable V. */
static bool
return !var_is_value_missing (v, &f->value, MV_ANY);
}
+struct freq_compare_aux
+ {
+ bool by_freq;
+ bool ascending_freq;
+
+ int width;
+ bool ascending_value;
+ };
+
+static int
+compare_freq (const void *a_, const void *b_, const void *aux_)
+{
+ const struct freq_compare_aux *aux = aux_;
+ const struct freq *a = a_;
+ const struct freq *b = b_;
+
+ if (aux->by_freq && a->count != b->count)
+ {
+ int cmp = a->count > b->count ? 1 : -1;
+ return aux->ascending_freq ? cmp : -cmp;
+ }
+ else
+ {
+ int cmp = value_compare_3way (&a->value, &b->value, aux->width);
+ return aux->ascending_value ? cmp : -cmp;
+ }
+}
/* Summarizes the frequency table data for variable V. */
static void
-postprocess_freq_tab (struct var_freqs *vf)
+postprocess_freq_tab (const struct frq_proc *frq, struct var_freqs *vf)
{
struct freq_tab *ft = &vf->tab;
- algo_compare_func *compare;
+ struct freq_compare_aux aux;
size_t count;
struct freq *freqs, *f;
size_t i;
ft->n_missing = count - ft->n_valid;
/* Sort data. */
- compare = get_freq_comparator (cmd.sort, var_get_type (vf->var));
- sort (ft->valid, ft->n_valid, sizeof *ft->valid, compare, vf);
- sort (ft->missing, ft->n_missing, sizeof *ft->missing, compare, vf);
+ aux.by_freq = frq->sort == FRQ_AFREQ || frq->sort == FRQ_DFREQ;
+ aux.ascending_freq = frq->sort != FRQ_DFREQ;
+ aux.width = vf->width;
+ aux.ascending_value = frq->sort != FRQ_DVALUE;
+ sort (ft->valid, ft->n_valid, sizeof *ft->valid, compare_freq, &aux);
+ sort (ft->missing, ft->n_missing, sizeof *ft->missing, compare_freq, &aux);
/* Summary statistics. */
ft->valid_cases = 0.0;
size_t n_vars;
size_t i;
- lex_match (lexer, '=');
+ lex_match (lexer, T_EQUALS);
if (lex_token (lexer) != T_ALL
&& (lex_token (lexer) != T_ID
- || dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) == NULL))
+ || dict_lookup_var (dataset_dict (ds), lex_tokcstr (lexer)) == NULL))
return 2;
/* Get list of current variables, to avoid duplicates. */
{
struct frq_proc *frq = frq_;
- lex_match (lexer, '=');
- if ((lex_token (lexer) == T_ID && dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) != NULL)
+ lex_match (lexer, T_EQUALS);
+ if ((lex_token (lexer) == T_ID
+ && dict_lookup_var (dataset_dict (ds), lex_tokcstr (lexer)) != NULL)
|| lex_token (lexer) == T_ID)
for (;;)
{
if (!parse_variables_const (lexer, dataset_dict (ds), &v, &n,
PV_NO_DUPLICATE | PV_NUMERIC))
return 0;
- if (lex_match (lexer, '('))
+ if (lex_match (lexer, T_LPAREN))
{
nl = ml = 0;
dl = NULL;
}
dl[nl++] = lex_tokval (lexer);
lex_get (lexer);
- lex_match (lexer, ',');
+ lex_match (lexer, T_COMMA);
}
/* Note that nl might still be 0 and dl might still be
NULL. That's okay. */
- if (!lex_match (lexer, ')'))
+ if (!lex_match (lexer, T_RPAREN))
{
free (v);
msg (SE, _("`)' expected after GROUPED interval list."));
}
free (v);
- if (!lex_match (lexer, '/'))
- break;
- if ((lex_token (lexer) != T_ID || dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) != NULL)
- && lex_token (lexer) != T_ALL)
- {
- lex_put_back (lexer, '/');
- break;
- }
+ if (lex_token (lexer) != T_SLASH)
+ break;
+
+ if ((lex_next_token (lexer, 1) == T_ID
+ && dict_lookup_var (dataset_dict (ds),
+ lex_next_tokcstr (lexer, 1)))
+ || lex_next_token (lexer, 1) == T_ALL)
+ {
+ /* The token after the slash is a variable name. Keep parsing. */
+ lex_get (lexer);
+ }
+ else
+ {
+ /* The token after the slash must be the start of a new
+ subcommand. Let the caller see the slash. */
+ break;
+ }
}
return 1;
order. If SHOW is true, the percentile will be shown in the statistics
box, otherwise it will be hidden. */
static void
-add_percentile (struct frq_proc *frq, double x, bool show)
+add_percentile (struct frq_proc *frq, double x, bool show,
+ size_t *allocated_percentiles)
{
int i;
- for (i = 0; i < n_percentiles; i++)
+ /* Do nothing if it's already in the list */
+ for (i = 0; i < frq->n_percentiles; i++)
{
- /* Do nothing if it's already in the list */
- if ( fabs(x - percentiles[i].p) < DBL_EPSILON )
+ struct percentile *pc = &frq->percentiles[i];
+
+ if ( fabs(x - pc->p) < DBL_EPSILON )
{
- if (show && !percentiles[i].show)
+ if (show && !pc->show)
{
- n_show_percentiles++;
- percentiles[i].show = true;
+ frq->n_show_percentiles++;
+ pc->show = true;
}
return;
}
- if (x < percentiles[i].p)
+ if (x < pc->p)
break;
}
- if (i >= n_percentiles || x != percentiles[i].p)
- {
- percentiles = pool_nrealloc (frq->pool, percentiles,
- n_percentiles + 1, sizeof *percentiles);
- insert_element (percentiles, n_percentiles, sizeof *percentiles, i);
- percentiles[i].p = x;
- percentiles[i].show = show;
- n_percentiles++;
- if (show)
- n_show_percentiles++;
- }
+ if (frq->n_percentiles >= *allocated_percentiles)
+ frq->percentiles = x2nrealloc (frq->percentiles, allocated_percentiles,
+ sizeof *frq->percentiles);
+ insert_element (frq->percentiles, frq->n_percentiles,
+ sizeof *frq->percentiles, i);
+ frq->percentiles[i].p = x;
+ frq->percentiles[i].show = show;
+ frq->n_percentiles++;
+ if (show)
+ frq->n_show_percentiles++;
}
/* Comparison functions. */
-/* Ascending numeric compare of values. */
-static int
-compare_value_numeric_a (const void *a_, const void *b_,
- const void *vf_ UNUSED)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
-
- if (a->value.f > b->value.f)
- return 1;
- else if (a->value.f < b->value.f)
- return -1;
- else
- return 0;
-}
-
-/* Ascending string compare of values. */
-static int
-compare_value_alpha_a (const void *a_, const void *b_, const void *vf_)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
- const struct var_freqs *vf = vf_;
-
- return value_compare_3way (&a->value, &b->value, vf->width);
-}
-
-/* Descending numeric compare of values. */
-static int
-compare_value_numeric_d (const void *a, const void *b, const void *vf_ UNUSED)
-{
- return -compare_value_numeric_a (a, b, vf_);
-}
-
-/* Descending string compare of values. */
-static int
-compare_value_alpha_d (const void *a, const void *b, const void *vf_)
-{
- return -compare_value_alpha_a (a, b, vf_);
-}
-
-/* Ascending numeric compare of frequency;
- secondary key on ascending numeric value. */
-static int
-compare_freq_numeric_a (const void *a_, const void *b_, const void *vf_ UNUSED)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
-
- if (a->count > b->count)
- return 1;
- else if (a->count < b->count)
- return -1;
-
- if (a->value.f > b->value.f)
- return 1;
- else if (a->value.f < b->value.f)
- return -1;
- else
- return 0;
-}
-
-/* Ascending numeric compare of frequency;
- secondary key on ascending string value. */
-static int
-compare_freq_alpha_a (const void *a_, const void *b_, const void *vf_)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
- const struct var_freqs *vf = vf_;
-
- if (a->count > b->count)
- return 1;
- else if (a->count < b->count)
- return -1;
- else
- return value_compare_3way (&a->value, &b->value, vf->width);
-}
-
-/* Descending numeric compare of frequency;
- secondary key on ascending numeric value. */
-static int
-compare_freq_numeric_d (const void *a_, const void *b_, const void *vf_ UNUSED)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
-
- if (a->count > b->count)
- return -1;
- else if (a->count < b->count)
- return 1;
-
- if (a->value.f > b->value.f)
- return 1;
- else if (a->value.f < b->value.f)
- return -1;
- else
- return 0;
-}
-
-/* Descending numeric compare of frequency;
- secondary key on ascending string value. */
-static int
-compare_freq_alpha_d (const void *a_, const void *b_, const void *vf_)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
- const struct var_freqs *vf = vf_;
-
- if (a->count > b->count)
- return -1;
- else if (a->count < b->count)
- return 1;
- else
- return value_compare_3way (&a->value, &b->value, vf->width);
-}
\f
/* Frequency table display. */
\f
/* Statistical display. */
-/* Calculates all the pertinent statistics for variable V, putting them in
- array D[]. */
+static double
+calc_percentile (double p, double valid_cases, double x1, double x2)
+{
+ double s, dummy;
+
+ s = (settings_get_algorithm () != COMPATIBLE
+ ? modf ((valid_cases - 1) * p, &dummy)
+ : modf ((valid_cases + 1) * p - 1, &dummy));
+
+ return x1 + (x2 - x1) * s;
+}
+
+/* Calculates all of the percentiles for VF within FRQ. */
static void
-calc_stats (const struct var_freqs *vf, double d[FRQ_N_STATS])
+calc_percentiles (const struct frq_proc *frq, const struct var_freqs *vf)
{
const struct freq_tab *ft = &vf->tab;
double W = ft->valid_cases;
- struct moments *m;
- struct freq *f=0;
- int most_often;
- double X_mode;
-
+ const struct freq *f;
+ int percentile_idx;
double rank;
- int i = 0;
- int idx;
-
- /* Calculate percentiles. */
assert (ft->n_valid > 0);
- for (i = 0; i < n_percentiles; i++)
- {
- percentiles[i].flag = 0;
- percentiles[i].flag2 = 0;
- }
-
rank = 0;
- for (idx = 0; idx < ft->n_valid; ++idx)
+ percentile_idx = 0;
+ for (f = ft->valid; f < ft->missing; f++)
{
- static double prev_value = SYSMIS;
- f = &ft->valid[idx];
- rank += f->count ;
- for (i = 0; i < n_percentiles; i++)
+ rank += f->count;
+ for (; percentile_idx < frq->n_percentiles; percentile_idx++)
{
- double tp;
- if ( percentiles[i].flag2 ) continue ;
-
- if ( settings_get_algorithm () != COMPATIBLE )
- tp =
- (ft->valid_cases - 1) * percentiles[i].p;
- else
- tp =
- (ft->valid_cases + 1) * percentiles[i].p - 1;
-
- if ( percentiles[i].flag )
- {
- percentiles[i].x2 = f->value.f;
- percentiles[i].x1 = prev_value;
- percentiles[i].flag2 = 1;
- continue;
- }
-
- if (rank > tp )
- {
- if ( f->count > 1 && rank - (f->count - 1) > tp )
- {
- percentiles[i].x2 = percentiles[i].x1 = f->value.f;
- percentiles[i].flag2 = 1;
- }
- else
- {
- percentiles[i].flag=1;
- }
+ struct percentile *pc = &frq->percentiles[percentile_idx];
+ double tp;
- continue;
- }
+ tp = (settings_get_algorithm () == ENHANCED
+ ? (W - 1) * pc->p
+ : (W + 1) * pc->p - 1);
+
+ if (rank <= tp)
+ break;
+
+ if (tp + 1 < rank || f + 1 >= ft->missing)
+ pc->value = f->value.f;
+ else
+ pc->value = calc_percentile (pc->p, W, f->value.f, f[1].value.f);
}
- prev_value = f->value.f;
}
-
- for (i = 0; i < n_percentiles; i++)
+ for (; percentile_idx < frq->n_percentiles; percentile_idx++)
{
- /* Catches the case when p == 100% */
- if ( ! percentiles[i].flag2 )
- percentiles[i].x1 = percentiles[i].x2 = f->value.f;
-
- /*
- printf("percentile %d (p==%.2f); X1 = %g; X2 = %g\n",
- i,percentiles[i].p,percentiles[i].x1,percentiles[i].x2);
- */
+ struct percentile *pc = &frq->percentiles[percentile_idx];
+ pc->value = ft->valid[ft->n_valid - 1].value.f;
}
+}
- for (i = 0; i < n_percentiles; i++)
- {
- double s;
-
- double dummy;
- if ( settings_get_algorithm () != COMPATIBLE )
- {
- s = modf((ft->valid_cases - 1) * percentiles[i].p , &dummy);
- }
- else
- {
- s = modf((ft->valid_cases + 1) * percentiles[i].p -1, &dummy);
- }
-
- percentiles[i].value = percentiles[i].x1 +
- ( percentiles[i].x2 - percentiles[i].x1) * s ;
- }
+/* Calculates all the pertinent statistics for VF, putting them in array
+ D[]. */
+static void
+calc_stats (const struct var_freqs *vf, double d[FRQ_N_STATS])
+{
+ const struct freq_tab *ft = &vf->tab;
+ double W = ft->valid_cases;
+ const struct freq *f;
+ struct moments *m;
+ int most_often;
+ double X_mode;
+ assert (ft->n_valid > 0);
/* Calculate the mode. */
most_often = -1;
/* Displays a table of all the statistics requested for variable V. */
static void
-dump_statistics (const struct var_freqs *vf, const struct variable *wv)
+dump_statistics (const struct frq_proc *frq, const struct var_freqs *vf,
+ const struct variable *wv)
{
const struct fmt_spec *wfmt = wv ? var_get_print_format (wv) : &F_8_0;
const struct freq_tab *ft = &vf->tab;
return;
}
calc_stats (vf, stat_value);
+ calc_percentiles (frq, vf);
- t = tab_create (3, n_stats + n_show_percentiles + 2);
+ t = tab_create (3, frq->n_stats + frq->n_show_percentiles + 2);
tab_box (t, TAL_1, TAL_1, -1, -1 , 0 , 0 , 2, tab_nr(t) - 1) ;
r=2; /* N missing and N valid are always dumped */
for (i = 0; i < FRQ_N_STATS; i++)
- if (stats & BIT_INDEX (i))
+ if (frq->stats & BIT_INDEX (i))
{
tab_text (t, 0, r, TAB_LEFT | TAT_TITLE,
gettext (st_name[i].s10));
tab_double (t, 2, 0, TAB_NONE, ft->valid_cases, wfmt);
tab_double (t, 2, 1, TAB_NONE, ft->total_cases - ft->valid_cases, wfmt);
- for (i = 0; i < n_percentiles; i++, r++)
+ for (i = 0; i < frq->n_percentiles; i++, r++)
{
- if (!percentiles[i].show)
+ struct percentile *pc = &frq->percentiles[i];
+
+ if (!pc->show)
continue;
if ( i == 0 )
tab_text (t, 0, r, TAB_LEFT | TAT_TITLE, _("Percentiles"));
}
- if (percentiles[i].p == 0.5)
+ if (pc->p == 0.5)
tab_text (t, 1, r, TAB_LEFT, _("50 (Median)"));
else
- tab_fixed (t, 1, r, TAB_LEFT, percentiles[i].p * 100, 3, 0);
- tab_double (t, 2, r, TAB_NONE, percentiles[i].value,
- var_get_print_format (vf->var));
+ tab_fixed (t, 1, r, TAB_LEFT, pc->p * 100, 3, 0);
+ tab_double (t, 2, r, TAB_NONE, pc->value,
+ var_get_print_format (vf->var));
}
tab_title (t, "%s", var_to_string (vf->var));
}
static double
-calculate_iqr (void)
+calculate_iqr (const struct frq_proc *frq)
{
double q1 = SYSMIS;
double q3 = SYSMIS;
int i;
- for (i = 0; i < n_percentiles; i++)
+ for (i = 0; i < frq->n_percentiles; i++)
{
- if (fabs (0.25 - percentiles[i].p) < DBL_EPSILON)
- q1 = percentiles[i].value;
- else if (fabs (0.75 - percentiles[i].p) < DBL_EPSILON)
- q3 = percentiles[i].value;
+ struct percentile *pc = &frq->percentiles[i];
+
+ if (fabs (0.25 - pc->p) < DBL_EPSILON)
+ q1 = pc->value;
+ else if (fabs (0.75 - pc->p) < DBL_EPSILON)
+ q3 = pc->value;
}
return q1 == SYSMIS || q3 == SYSMIS ? SYSMIS : q3 - q1;
/* Create a gsl_histogram from a freq_tab */
struct histogram *
-freq_tab_to_hist (const struct freq_tab *ft, const struct variable *var)
+freq_tab_to_hist (const struct frq_proc *frq, const struct freq_tab *ft,
+ const struct variable *var)
{
double x_min, x_max, valid_freq;
int i;
valid_freq = 0;
for (i = 0; i < ft->n_valid; i++)
{
- const struct freq *frq = &ft->valid[i];
- if (chart_includes_value (&hist, var, &frq->value))
+ const struct freq *f = &ft->valid[i];
+ if (chart_includes_value (frq->hist, var, &f->value))
{
- x_min = MIN (x_min, frq->value.f);
- x_max = MAX (x_max, frq->value.f);
- valid_freq += frq->count;
+ x_min = MIN (x_min, f->value.f);
+ x_max = MAX (x_max, f->value.f);
+ valid_freq += f->count;
}
}
/* Freedman-Diaconis' choice of bin width. */
- iqr = calculate_iqr ();
+ iqr = calculate_iqr (frq);
if (iqr != SYSMIS)
{
double bin_width = 2 * iqr / pow (valid_freq, 1.0 / 3.0);
histogram = histogram_create (bins, x_min, x_max);
for (i = 0; i < ft->n_valid; i++)
{
- const struct freq *frq = &ft->valid[i];
- if (chart_includes_value (&hist, var, &frq->value))
- histogram_add (histogram, frq->value.f, frq->count);
+ const struct freq *f = &ft->valid[i];
+ if (chart_includes_value (frq->hist, var, &f->value))
+ histogram_add (histogram, f->value.f, f->count);
}
return histogram;
}
static int
-add_slice (const struct freq *freq, const struct variable *var,
- struct slice *slice)
+add_slice (const struct frq_chart *pie, const struct freq *freq,
+ const struct variable *var, struct slice *slice)
{
- if (chart_includes_value (&pie, var, &freq->value))
+ if (chart_includes_value (pie, var, &freq->value))
{
ds_init_empty (&slice->label);
var_append_value_name (var, &freq->value, &slice->label);
The caller is responsible for freeing slices
*/
static struct slice *
-freq_tab_to_slice_array(const struct freq_tab *frq_tab,
+freq_tab_to_slice_array(const struct frq_chart *pie,
+ const struct freq_tab *frq_tab,
const struct variable *var,
int *n_slicesp)
{
n_slices = 0;
for (i = 0; i < frq_tab->n_valid; i++)
- n_slices += add_slice (&frq_tab->valid[i], var, &slices[n_slices]);
+ n_slices += add_slice (pie, &frq_tab->valid[i], var, &slices[n_slices]);
for (i = 0; i < frq_tab->n_missing; i++)
- n_slices += add_slice (&frq_tab->missing[i], var, &slices[n_slices]);
+ n_slices += add_slice (pie, &frq_tab->missing[i], var, &slices[n_slices]);
*n_slicesp = n_slices;
return slices;
static void
-do_piechart(const struct variable *var, const struct freq_tab *frq_tab)
+do_piechart(const struct frq_chart *pie, const struct variable *var,
+ const struct freq_tab *frq_tab)
{
struct slice *slices;
int n_slices, i;
- slices = freq_tab_to_slice_array (frq_tab, var, &n_slices);
+ slices = freq_tab_to_slice_array (pie, frq_tab, var, &n_slices);
if (n_slices < 2)
msg (SW, _("Omitting pie chart for %s, which has only %d unique values."),
projects / pspp-builds.git / blobdiff