/* PSPP - a program for statistical analysis.
- Copyright (C) 2009 Free Software Foundation, Inc.
+ Copyright (C) 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 <config.h>
-#include <data/procedure.h>
-#include <language/lexer/variable-parser.h>
-#include <language/lexer/value-parser.h>
-#include <language/command.h>
-#include <language/lexer/lexer.h>
-
-#include <data/casegrouper.h>
-#include <data/casereader.h>
-#include <data/casewriter.h>
-#include <data/dictionary.h>
-#include <data/format.h>
-#include <math/sort.h>
-#include <data/subcase.h>
-
-
-#include <libpspp/misc.h>
+#include "language/stats/roc.h"
#include <gsl/gsl_cdf.h>
-#include <output/table.h>
-#include <output/charts/plot-chart.h>
-#include <output/charts/cartesian.h>
+#include "data/casegrouper.h"
+#include "data/casereader.h"
+#include "data/casewriter.h"
+#include "data/dataset.h"
+#include "data/dictionary.h"
+#include "data/format.h"
+#include "data/subcase.h"
+#include "language/command.h"
+#include "language/lexer/lexer.h"
+#include "language/lexer/value-parser.h"
+#include "language/lexer/variable-parser.h"
+#include "libpspp/misc.h"
+#include "math/sort.h"
+#include "output/chart-item.h"
+#include "output/charts/roc-chart.h"
+#include "output/tab.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
const struct variable **vars;
const struct dictionary *dict;
- const struct variable *state_var ;
+ const struct variable *state_var;
union value state_value;
/* Plot the roc curve */
roc.pos = roc.pos_weighted = 0;
roc.neg = roc.neg_weighted = 0;
roc.dict = dataset_dict (ds);
+ roc.state_var = NULL;
+ lex_match (lexer, T_SLASH);
if (!parse_variables_const (lexer, dict, &roc.vars, &roc.n_vars,
PV_APPEND | PV_NO_DUPLICATE | PV_NUMERIC))
goto error;
roc.state_var = parse_variable (lexer, dict);
- if ( !lex_force_match (lexer, '('))
+ if ( !lex_force_match (lexer, T_LPAREN))
{
goto error;
}
+ value_init (&roc.state_value, var_get_width (roc.state_var));
parse_value (lexer, &roc.state_value, var_get_width (roc.state_var));
- if ( !lex_force_match (lexer, ')'))
+ if ( !lex_force_match (lexer, T_RPAREN))
{
goto error;
}
- while (lex_token (lexer) != '.')
+ while (lex_token (lexer) != T_ENDCMD)
{
- lex_match (lexer, '/');
+ lex_match (lexer, T_SLASH);
if (lex_match_id (lexer, "MISSING"))
{
- lex_match (lexer, '=');
- while (lex_token (lexer) != '.' && lex_token (lexer) != '/')
+ lex_match (lexer, T_EQUALS);
+ while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH)
{
if (lex_match_id (lexer, "INCLUDE"))
{
}
else if (lex_match_id (lexer, "PLOT"))
{
- lex_match (lexer, '=');
+ lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "CURVE"))
{
roc.curve = true;
- if (lex_match (lexer, '('))
+ if (lex_match (lexer, T_LPAREN))
{
roc.reference = true;
lex_force_match_id (lexer, "REFERENCE");
- lex_force_match (lexer, ')');
+ lex_force_match (lexer, T_RPAREN);
}
}
else if (lex_match_id (lexer, "NONE"))
}
else if (lex_match_id (lexer, "PRINT"))
{
- lex_match (lexer, '=');
- while (lex_token (lexer) != '.' && lex_token (lexer) != '/')
+ lex_match (lexer, T_EQUALS);
+ while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH)
{
if (lex_match_id (lexer, "SE"))
{
}
else if (lex_match_id (lexer, "CRITERIA"))
{
- lex_match (lexer, '=');
- while (lex_token (lexer) != '.' && lex_token (lexer) != '/')
+ lex_match (lexer, T_EQUALS);
+ while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH)
{
if (lex_match_id (lexer, "CUTOFF"))
{
- lex_force_match (lexer, '(');
+ lex_force_match (lexer, T_LPAREN);
if (lex_match_id (lexer, "INCLUDE"))
{
roc.exclude = MV_SYSTEM;
lex_error (lexer, NULL);
goto error;
}
- lex_force_match (lexer, ')');
+ lex_force_match (lexer, T_RPAREN);
}
else if (lex_match_id (lexer, "TESTPOS"))
{
- lex_force_match (lexer, '(');
+ lex_force_match (lexer, T_LPAREN);
if (lex_match_id (lexer, "LARGE"))
{
roc.invert = false;
lex_error (lexer, NULL);
goto error;
}
- lex_force_match (lexer, ')');
+ lex_force_match (lexer, T_RPAREN);
}
else if (lex_match_id (lexer, "CI"))
{
- lex_force_match (lexer, '(');
+ lex_force_match (lexer, T_LPAREN);
lex_force_num (lexer);
roc.ci = lex_number (lexer);
lex_get (lexer);
- lex_force_match (lexer, ')');
+ lex_force_match (lexer, T_RPAREN);
}
else if (lex_match_id (lexer, "DISTRIBUTION"))
{
- lex_force_match (lexer, '(');
+ lex_force_match (lexer, T_LPAREN);
if (lex_match_id (lexer, "FREE"))
{
roc.bi_neg_exp = false;
lex_error (lexer, NULL);
goto error;
}
- lex_force_match (lexer, ')');
+ lex_force_match (lexer, T_RPAREN);
}
else
{
if ( ! run_roc (ds, &roc))
goto error;
+ value_destroy (&roc.state_value, var_get_width (roc.state_var));
free (roc.vars);
return CMD_SUCCESS;
error:
+ if ( roc.state_var)
+ value_destroy (&roc.state_value, var_get_width (roc.state_var));
free (roc.vars);
return CMD_FAILURE;
}
double max;
};
-#define CUTPOINT 0
-#define TP 1
-#define FN 2
-#define TN 3
-#define FP 4
-
-
/*
Return a new casereader based upon CUTPOINT_RDR.
The number of "positive" cases are placed into
WEIGHT is the value of a single count.
*/
static struct casereader *
-accumulate_counts (struct casereader *cutpoint_rdr,
+accumulate_counts (struct casereader *input,
double result, double weight,
bool (*pos_cond) (double, double),
int true_index, int false_index)
{
- const struct caseproto *proto = casereader_get_proto (cutpoint_rdr);
+ const struct caseproto *proto = casereader_get_proto (input);
struct casewriter *w =
autopaging_writer_create (proto);
- struct casereader *r = casereader_clone (cutpoint_rdr);
struct ccase *cpc;
double prev_cp = SYSMIS;
- for ( ; (cpc = casereader_read (r) ); case_unref (cpc))
+ for ( ; (cpc = casereader_read (input) ); case_unref (cpc))
{
struct ccase *new_case;
- const double cp = case_data_idx (cpc, CUTPOINT)->f;
+ const double cp = case_data_idx (cpc, ROC_CUTPOINT)->f;
assert (cp != SYSMIS);
casewriter_write (w, new_case);
}
- casereader_destroy (r);
+ casereader_destroy (input);
return casewriter_make_reader (w);
}
casereader_destroy (r2);
}
+
casereader_destroy (r1);
casereader_destroy (rclone);
+ caseproto_unref (proto);
+
return casewriter_make_reader (wtr);
}
return process_group (var, reader, gt, dict, &rs->n1,
&rs->cutpoint_rdr,
ge,
- TP, FN);
+ ROC_TP, ROC_FN);
}
/*
return process_group (var, reader, lt, dict, &rs->n2,
&rs->cutpoint_rdr,
lt,
- TN, FP);
+ ROC_TN, ROC_FP);
}
{
struct ccase *cc = case_create (casewriter_get_proto (writer));
- case_data_rw_idx (cc, CUTPOINT)->f = cutpoint;
- case_data_rw_idx (cc, TP)->f = 0;
- case_data_rw_idx (cc, FN)->f = 0;
- case_data_rw_idx (cc, TN)->f = 0;
- case_data_rw_idx (cc, FP)->f = 0;
+ case_data_rw_idx (cc, ROC_CUTPOINT)->f = cutpoint;
+ case_data_rw_idx (cc, ROC_TP)->f = 0;
+ case_data_rw_idx (cc, ROC_FN)->f = 0;
+ case_data_rw_idx (cc, ROC_TN)->f = 0;
+ case_data_rw_idx (cc, ROC_FP)->f = 0;
casewriter_write (writer, cc);
}
/*
Create and initialise the rs[x].cutpoint_rdr casereaders. That is, the readers will
- be created with width 5, ready to take the values (cutpoint, TP, FN, TN, FP), and the
+ be created with width 5, ready to take the values (cutpoint, ROC_TP, ROC_FN, ROC_TN, ROC_FP), and the
reader will be populated with its final number of cases.
- However on exit from this function, only CUTPOINT entries will be set to their final
+ However on exit from this function, only ROC_CUTPOINT entries will be set to their final
value. The other entries will be initialised to zero.
*/
static void
int i;
struct casereader *r = casereader_clone (input);
struct ccase *c;
- struct caseproto *proto = caseproto_create ();
- struct subcase ordering;
- subcase_init (&ordering, CUTPOINT, 0, SC_ASCEND);
+ {
+ struct caseproto *proto = caseproto_create ();
+ struct subcase ordering;
+ subcase_init (&ordering, ROC_CUTPOINT, 0, SC_ASCEND);
- proto = caseproto_add_width (proto, 0); /* cutpoint */
- proto = caseproto_add_width (proto, 0); /* TP */
- proto = caseproto_add_width (proto, 0); /* FN */
- proto = caseproto_add_width (proto, 0); /* TN */
- proto = caseproto_add_width (proto, 0); /* FP */
+ proto = caseproto_add_width (proto, 0); /* cutpoint */
+ proto = caseproto_add_width (proto, 0); /* ROC_TP */
+ proto = caseproto_add_width (proto, 0); /* ROC_FN */
+ proto = caseproto_add_width (proto, 0); /* ROC_TN */
+ proto = caseproto_add_width (proto, 0); /* ROC_FP */
- for (i = 0 ; i < roc->n_vars; ++i)
- {
- rs[i].cutpoint_wtr = sort_create_writer (&ordering, proto);
- rs[i].prev_result = SYSMIS;
- rs[i].max = -DBL_MAX;
- rs[i].min = DBL_MAX;
- }
+ for (i = 0 ; i < roc->n_vars; ++i)
+ {
+ rs[i].cutpoint_wtr = sort_create_writer (&ordering, proto);
+ rs[i].prev_result = SYSMIS;
+ rs[i].max = -DBL_MAX;
+ rs[i].min = DBL_MAX;
+ }
+
+ caseproto_unref (proto);
+ subcase_destroy (&ordering);
+ }
for (; (c = casereader_read (r)) != NULL; case_unref (c))
{
struct casereader *negatives = NULL;
struct casereader *positives = NULL;
- struct caseproto *n_proto = caseproto_create ();
+ struct caseproto *n_proto = NULL;
struct subcase up_ordering;
struct subcase down_ordering;
struct ccase *c;
struct ccase *cpos;
- struct casereader *n_neg ;
+ struct casereader *n_neg_reader ;
const struct variable *var = roc->vars[i];
struct casereader *neg ;
struct casereader *pos = casereader_clone (positives);
-
- struct casereader *n_pos =
+ struct casereader *n_pos_reader =
process_positive_group (var, pos, dict, &rs[i]);
if ( negatives == NULL)
neg = casereader_clone (negatives);
- n_neg = process_negative_group (var, neg, dict, &rs[i]);
-
+ n_neg_reader = process_negative_group (var, neg, dict, &rs[i]);
/* Merge the n_pos and n_neg casereaders */
w = sort_create_writer (&up_ordering, n_proto);
- for ( ; (cpos = casereader_read (n_pos) ); case_unref (cpos))
+ for ( ; (cpos = casereader_read (n_pos_reader) ); case_unref (cpos))
{
struct ccase *pos_case = case_create (n_proto);
struct ccase *cneg;
const double jpos = case_data_idx (cpos, VALUE)->f;
- while ((cneg = casereader_read (n_neg)))
+ while ((cneg = casereader_read (n_neg_reader)))
{
struct ccase *nc = case_create (n_proto);
casewriter_write (w, pos_case);
}
+ casereader_destroy (n_pos_reader);
+ casereader_destroy (n_neg_reader);
+
/* These aren't used anymore */
#undef N_EQ
#undef N_PRED
prev_pos_gt = n_pos_gt;
}
+ casereader_destroy (r);
r = casewriter_make_reader (w);
}
prev_neg_lt = n_neg_lt;
}
+ casereader_destroy (r);
r = casewriter_make_reader (w);
}
struct ccase *prev_case = NULL;
for ( ; (c = casereader_read (r) ); case_unref (c))
{
- const struct ccase *next_case = casereader_peek (r, 0);
+ struct ccase *next_case = casereader_peek (r, 0);
const double j = case_data_idx (c, VALUE)->f;
double n_pos_eq = case_data_idx (c, N_POS_EQ)->f;
}
+ case_unref (next_case);
case_unref (prev_case);
prev_case = case_clone (c);
}
+ casereader_destroy (r);
+ case_unref (prev_case);
rs[i].auc /= rs[i].n1 * rs[i].n2;
if ( roc->invert )
casereader_destroy (positives);
casereader_destroy (negatives);
+ caseproto_unref (n_proto);
+ subcase_destroy (&up_ordering);
+ subcase_destroy (&down_ordering);
+
output_roc (rs, roc);
+
+ for (i = 0 ; i < roc->n_vars; ++i)
+ casereader_destroy (rs[i].cutpoint_rdr);
free (rs);
}
const int n_fields = roc->print_se ? 5 : 1;
const int n_cols = roc->n_vars > 1 ? n_fields + 1: n_fields;
const int n_rows = 2 + roc->n_vars;
- struct tab_table *tbl = tab_create (n_cols, n_rows, 0);
+ struct tab_table *tbl = tab_create (n_cols, n_rows);
if ( roc->n_vars > 1)
tab_title (tbl, _("Area Under the Curve"));
tab_headers (tbl, n_cols - n_fields, 0, 1, 0);
- tab_dim (tbl, tab_natural_dimensions, NULL);
tab_text (tbl, n_cols - n_fields, 1, TAT_TITLE, _("Area"));
tab_text (tbl, n_cols - 2, 1, TAT_TITLE, _("Lower Bound"));
tab_text (tbl, n_cols - 1, 1, TAT_TITLE, _("Upper Bound"));
- tab_joint_text (tbl, n_cols - 2, 0, 4, 0,
- TAT_TITLE | TAB_CENTER | TAT_PRINTF,
- _("Asymp. %g%% Confidence Interval"), roc->ci);
+ tab_joint_text_format (tbl, n_cols - 2, 0, 4, 0,
+ TAT_TITLE | TAB_CENTER,
+ _("Asymp. %g%% Confidence Interval"), roc->ci);
tab_vline (tbl, 0, n_cols - 1, 0, 0);
tab_hline (tbl, TAL_1, n_cols - 2, n_cols - 1, 1);
}
{
const int n_cols = 3;
const int n_rows = 4;
- struct tab_table *tbl = tab_create (n_cols, n_rows, 0);
+ struct tab_table *tbl = tab_create (n_cols, n_rows);
tab_title (tbl, _("Case Summary"));
tab_headers (tbl, 1, 0, 2, 0);
- tab_dim (tbl, tab_natural_dimensions, NULL);
-
tab_box (tbl,
TAL_2, TAL_2,
-1, -1,
for (i = 0; i < roc->n_vars; ++i)
n_rows += casereader_count_cases (rs[i].cutpoint_rdr);
- tbl = tab_create (n_cols, n_rows, 0);
+ tbl = tab_create (n_cols, n_rows);
if ( roc->n_vars > 1)
tab_title (tbl, _("Coordinates of the Curve"));
tab_headers (tbl, 1, 0, 1, 0);
- tab_dim (tbl, tab_natural_dimensions, NULL);
-
tab_hline (tbl, TAL_2, 0, n_cols - 1, 1);
if ( roc->n_vars > 1)
for (; (cc = casereader_read (r)) != NULL;
case_unref (cc), x++)
{
- const double se = case_data_idx (cc, TP)->f /
+ const double se = case_data_idx (cc, ROC_TP)->f /
(
- case_data_idx (cc, TP)->f
+ case_data_idx (cc, ROC_TP)->f
+
- case_data_idx (cc, FN)->f
+ case_data_idx (cc, ROC_FN)->f
);
- const double sp = case_data_idx (cc, TN)->f /
+ const double sp = case_data_idx (cc, ROC_TN)->f /
(
- case_data_idx (cc, TN)->f
+ case_data_idx (cc, ROC_TN)->f
+
- case_data_idx (cc, FP)->f
+ case_data_idx (cc, ROC_FP)->f
);
- tab_double (tbl, n_cols - 3, x, 0, case_data_idx (cc, CUTPOINT)->f,
+ tab_double (tbl, n_cols - 3, x, 0, case_data_idx (cc, ROC_CUTPOINT)->f,
var_get_print_format (roc->vars[i]));
tab_double (tbl, n_cols - 2, x, 0, se, NULL);
}
-static void
-draw_roc (struct roc_state *rs, const struct cmd_roc *roc)
-{
- int i;
-
- struct chart *roc_chart = chart_create ();
-
- chart_write_title (roc_chart, _("ROC Curve"));
- chart_write_xlabel (roc_chart, _("1 - Specificity"));
- chart_write_ylabel (roc_chart, _("Sensitivity"));
-
- chart_write_xscale (roc_chart, 0, 1, 5);
- chart_write_yscale (roc_chart, 0, 1, 5);
-
- if ( roc->reference )
- {
- chart_line (roc_chart, 1.0, 0,
- 0.0, 1.0,
- CHART_DIM_X);
- }
-
- for (i = 0; i < roc->n_vars; ++i)
- {
- struct ccase *cc;
- struct casereader *r = casereader_clone (rs[i].cutpoint_rdr);
-
- chart_vector_start (roc_chart, var_get_name (roc->vars[i]));
- for (; (cc = casereader_read (r)) != NULL;
- case_unref (cc))
- {
- double se = case_data_idx (cc, TP)->f;
- double sp = case_data_idx (cc, TN)->f;
-
- se /= case_data_idx (cc, FN)->f +
- case_data_idx (cc, TP)->f ;
-
- sp /= case_data_idx (cc, TN)->f +
- case_data_idx (cc, FP)->f ;
-
- chart_vector (roc_chart, 1 - sp, se);
- }
- chart_vector_end (roc_chart);
- casereader_destroy (r);
- }
-
- chart_write_legend (roc_chart);
-
- chart_submit (roc_chart);
-}
-
-
static void
output_roc (struct roc_state *rs, const struct cmd_roc *roc)
{
show_summary (roc);
if ( roc->curve )
- draw_roc (rs, roc);
+ {
+ struct roc_chart *rc;
+ size_t i;
+
+ rc = roc_chart_create (roc->reference);
+ for (i = 0; i < roc->n_vars; i++)
+ roc_chart_add_var (rc, var_get_name (roc->vars[i]),
+ rs[i].cutpoint_rdr);
+ roc_chart_submit (rc);
+ }
show_auc (rs, roc);
-
if ( roc->print_coords )
show_coords (rs, roc);
}
projects / pspp-builds.git / blobdiff