X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Fbinomial.c;h=f2ae51513eb7c4366c993d0875eb87fa10eef936;hb=033449d465fb54526d3984d55772e6a38e7d922c;hp=d9209f1bf0499dbcf8e78f8bb9ba1b6e27861683;hpb=83c4d7ff1c815395eb050aa6b16b6b740595cc92;p=pspp diff --git a/src/language/stats/binomial.c b/src/language/stats/binomial.c index d9209f1bf0..f2ae51513e 100644 --- a/src/language/stats/binomial.c +++ b/src/language/stats/binomial.c @@ -1,5 +1,5 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 2006 Free Software Foundation, Inc. + Copyright (C) 2006, 2009, 2010, 2011, 2014 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 @@ -15,36 +15,34 @@ along with this program. If not, see . */ #include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include - -#include "binomial.h" -#include "freq.h" - -#include "xalloc.h" - -#include "gettext.h" -#define _(msgid) gettext (msgid) - -#include +#include "language/stats/binomial.h" +#include #include #include -#include +#include "data/case.h" +#include "data/casereader.h" +#include "data/dataset.h" +#include "data/dictionary.h" +#include "data/format.h" +#include "data/value-labels.h" +#include "data/value.h" +#include "data/variable.h" +#include "language/stats/freq.h" +#include "libpspp/assertion.h" +#include "libpspp/compiler.h" +#include "libpspp/message.h" +#include "libpspp/misc.h" +#include "output/pivot-table.h" + +#include "gl/xalloc.h" +#include "gl/minmax.h" -#include +#include "gettext.h" +#define N_(msgid) msgid +#define _(msgid) gettext (msgid) static double calculate_binomial_internal (double n1, double n2, double p); @@ -62,8 +60,8 @@ static double calculate_binomial (double n1, double n2, double p) { const double n = n1 + n2; - const bool test_reversed = (n1 / n > p ) ; - if ( test_reversed ) + const bool test_reversed = (n1 / n > p) ; + if (test_reversed) { p = 1 - p ; swap (&n1, &n2); @@ -80,7 +78,7 @@ calculate_binomial_internal (double n1, double n2, double p) double sig1tailed = gsl_cdf_binomial_P (n1, p, n1 + n2); - if ( p == 0.5 ) + if (p == 0.5) return sig1tailed > 0.5 ? 1.0 :sig1tailed * 2.0; return sig1tailed ; @@ -89,50 +87,57 @@ calculate_binomial_internal (double n1, double n2, double p) static bool do_binomial (const struct dictionary *dict, struct casereader *input, - const struct binomial_test *bst, - struct freq_mutable *cat1, - struct freq_mutable *cat2, + const struct one_sample_test *ost, + struct freq *cat1, + struct freq *cat2, enum mv_class exclude - ) + ) { + const struct binomial_test *bst = UP_CAST (ost, const struct binomial_test, parent); bool warn = true; - const struct one_sample_test *ost = (const struct one_sample_test *) bst; - struct ccase c; + struct ccase *c; - while (casereader_read(input, &c)) + for (; (c = casereader_read (input)) != NULL; case_unref (c)) { int v; - double w = dict_get_case_weight (dict, &c, &warn); + double w = dict_get_case_weight (dict, c, &warn); - for (v = 0 ; v < ost->n_vars ; ++v ) + for (v = 0 ; v < ost->n_vars ; ++v) { const struct variable *var = ost->vars[v]; - const union value *value = case_data (&c, var); - int width = var_get_width (var); + double value = case_num (c, var); - if (var_is_value_missing (var, value, exclude)) - break; + if (var_is_num_missing (var, value) & exclude) + continue; - if ( NULL == cat1[v].value ) + if (bst->cutpoint != SYSMIS) { - cat1[v].value = value_dup (value, width); - cat1[v].count = w; + if (cat1[v].values[0].f >= value) + cat1[v].count += w; + else + cat2[v].count += w; } - else if ( 0 == compare_values (cat1[v].value, value, width)) - cat1[v].count += w; - else if ( NULL == cat2[v].value ) + else { - cat2[v].value = value_dup (value, width); - cat2[v].count = w; + if (SYSMIS == cat1[v].values[0].f) + { + cat1[v].values[0].f = value; + cat1[v].count = w; + } + else if (cat1[v].values[0].f == value) + cat1[v].count += w; + else if (SYSMIS == cat2[v].values[0].f) + { + cat2[v].values[0].f = value; + cat2[v].count = w; + } + else if (cat2[v].values[0].f == value) + cat2[v].count += w; + else if (bst->category1 == SYSMIS) + msg (ME, _("Variable %s is not dichotomous"), var_get_name (var)); } - else if ( 0 == compare_values (cat2[v].value, value, width)) - cat2[v].count += w; - else if ( bst->category1 == SYSMIS) - msg (ME, _("Variable %s is not dichotomous"), var_get_name (var)); } - - case_destroy (&c); } return casereader_destroy (input); } @@ -143,113 +148,109 @@ void binomial_execute (const struct dataset *ds, struct casereader *input, enum mv_class exclude, - const struct npar_test *test) + const struct npar_test *test, + bool exact UNUSED, + double timer UNUSED) { - int v; - const struct binomial_test *bst = (const struct binomial_test *) test; - const struct one_sample_test *ost = (const struct one_sample_test*) test; - - struct freq_mutable *cat1 = xzalloc (sizeof (*cat1) * ost->n_vars); - struct freq_mutable *cat2 = xzalloc (sizeof (*cat1) * ost->n_vars); + const struct dictionary *dict = dataset_dict (ds); + const struct one_sample_test *ost = UP_CAST (test, const struct one_sample_test, parent); + const struct binomial_test *bst = UP_CAST (ost, const struct binomial_test, parent); - assert ((bst->category1 == SYSMIS) == (bst->category2 == SYSMIS) ); + struct freq *cat[2]; + int i; - if ( bst->category1 != SYSMIS ) - { - union value v; - v.f = bst->category1; - cat1->value = value_dup (&v, 0); - } + assert ((bst->category1 == SYSMIS) == (bst->category2 == SYSMIS) || bst->cutpoint != SYSMIS); - if ( bst->category2 != SYSMIS ) + for (i = 0; i < 2; i++) { - union value v; - v.f = bst->category2; - cat2->value = value_dup (&v, 0); + double value; + if (i == 0) + value = bst->cutpoint != SYSMIS ? bst->cutpoint : bst->category1; + else + value = bst->category2; + + cat[i] = xnmalloc (ost->n_vars, sizeof *cat[i]); + for (size_t v = 0; v < ost->n_vars; v++) + { + cat[i][v].values[0].f = value; + cat[i][v].count = 0; + } } - if (do_binomial (dataset_dict(ds), input, bst, cat1, cat2, exclude)) + if (do_binomial (dataset_dict (ds), input, ost, cat[0], cat[1], exclude)) { - struct tab_table *table = tab_create (7, ost->n_vars * 3 + 1, 0); - - tab_dim (table, tab_natural_dimensions); - - tab_title (table, _("Binomial Test")); - - tab_headers (table, 2, 0, 1, 0); - - tab_box (table, TAL_1, TAL_1, -1, TAL_1, - 0, 0, table->nc - 1, tab_nr(table) - 1 ); - - for (v = 0 ; v < ost->n_vars; ++v) + struct pivot_table *table = pivot_table_create (N_("Binomial Test")); + pivot_table_set_weight_var (table, dict_get_weight (dict)); + + pivot_dimension_create ( + table, PIVOT_AXIS_COLUMN, N_("Statistics"), + N_("Category"), + N_("N"), PIVOT_RC_COUNT, + N_("Observed Prop."), PIVOT_RC_OTHER, + N_("Test Prop."), PIVOT_RC_OTHER, + (bst->p == 0.5 + ? N_("Exact Sig. (2-tailed)") + : N_("Exact Sig. (1-tailed)")), PIVOT_RC_SIGNIFICANCE); + + pivot_dimension_create (table, PIVOT_AXIS_ROW, N_("Groups"), + N_("Group 1"), N_("Group 2"), N_("Total")); + + struct pivot_dimension *variables = pivot_dimension_create ( + table, PIVOT_AXIS_ROW, N_("Variables")); + + for (size_t v = 0; v < ost->n_vars; ++v) { - double n_total, sig; - struct string catstr1; - struct string catstr2; const struct variable *var = ost->vars[v]; - ds_init_empty (&catstr1); - ds_init_empty (&catstr2); - - var_append_value_name (var, cat1[v].value, &catstr1); - var_append_value_name (var, cat2[v].value, &catstr2); - - tab_hline (table, TAL_1, 0, tab_nc (table) -1, 1 + v * 3); - - /* Titles */ - tab_text (table, 0, 1 + v * 3, TAB_LEFT, var_to_string (var)); - tab_text (table, 1, 1 + v * 3, TAB_LEFT, _("Group1")); - tab_text (table, 1, 2 + v * 3, TAB_LEFT, _("Group2")); - tab_text (table, 1, 3 + v * 3, TAB_LEFT, _("Total")); - - /* Test Prop */ - tab_float (table, 5, 1 + v * 3, TAB_NONE, bst->p, 8, 3); - - /* Category labels */ - tab_text (table, 2, 1 + v * 3, TAB_NONE, ds_cstr (&catstr1)); - tab_text (table, 2, 2 + v * 3, TAB_NONE, ds_cstr (&catstr2)); - - /* Observed N */ - tab_float (table, 3, 1 + v * 3, TAB_NONE, cat1[v].count, 8, 0); - tab_float (table, 3, 2 + v * 3, TAB_NONE, cat2[v].count, 8, 0); - - n_total = cat1[v].count + cat2[v].count; - tab_float (table, 3, 3 + v * 3, TAB_NONE, n_total, 8, 0); - - /* Observed Proportions */ - tab_float (table, 4, 1 + v * 3, TAB_NONE, - cat1[v].count / n_total, 8, 3); - tab_float (table, 4, 2 + v * 3, TAB_NONE, - cat2[v].count / n_total, 8, 3); - tab_float (table, 4, 3 + v * 3, TAB_NONE, - (cat1[v].count + cat2[v].count) / n_total, 8, 2); - - /* Significance */ - sig = calculate_binomial (cat1[v].count, cat2[v].count, bst->p); - tab_float (table, 6, 1 + v * 3, TAB_NONE, sig, 8, 3); - - ds_destroy (&catstr1); - ds_destroy (&catstr2); + int var_idx = pivot_category_create_leaf ( + variables->root, pivot_value_new_variable (var)); + + /* Category. */ + if (bst->cutpoint != SYSMIS) + pivot_table_put3 ( + table, 0, 0, var_idx, + pivot_value_new_user_text_nocopy ( + xasprintf ("<= %.*g", DBL_DIG + 1, bst->cutpoint))); + else + for (int i = 0; i < 2; i++) + pivot_table_put3 ( + table, 0, i, var_idx, + pivot_value_new_var_value (var, cat[i][v].values)); + + double n_total = cat[0][v].count + cat[1][v].count; + double sig = calculate_binomial (cat[0][v].count, cat[1][v].count, + bst->p); + struct entry + { + int stat_idx; + int group_idx; + double x; + } + entries[] = { + /* N. */ + { 1, 0, cat[0][v].count }, + { 1, 1, cat[1][v].count }, + { 1, 2, n_total }, + /* Observed Prop. */ + { 2, 0, cat[0][v].count / n_total }, + { 2, 1, cat[1][v].count / n_total }, + { 2, 2, 1.0 }, + /* Test Prop. */ + { 3, 0, bst->p }, + /* Significance. */ + { 4, 0, sig } + }; + for (size_t i = 0; i < sizeof entries / sizeof *entries; i++) + { + const struct entry *e = &entries[i]; + pivot_table_put3 (table, e->stat_idx, e->group_idx, + var_idx, pivot_value_new_number (e->x)); + } } - tab_text (table, 2, 0, TAB_CENTER, _("Category")); - tab_text (table, 3, 0, TAB_CENTER, _("N")); - tab_text (table, 4, 0, TAB_CENTER, _("Observed Prop.")); - tab_text (table, 5, 0, TAB_CENTER, _("Test Prop.")); - - tab_text (table, 6, 0, TAB_CENTER | TAT_PRINTF, - _("Exact Sig. (%d-tailed)"), - bst->p == 0.5 ? 2: 1); - - tab_vline (table, TAL_2, 2, 0, tab_nr (table) -1); - tab_submit (table); + pivot_table_submit (table); } - for (v = 0; v < ost->n_vars; v++) - { - free (cat1[v].value); - free (cat2[v].value); - } - free (cat1); - free (cat2); + for (i = 0; i < 2; i++) + free (cat[i]); }