--- /dev/null
+/* PSPP - a program for statistical analysis.
+ Copyright (C) 2011, 2020 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
+ the Free Software Foundation, either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <config.h>
+
+#include "t-test.h"
+
+#include <gsl/gsl_cdf.h>
+#include <math.h>
+#include "libpspp/misc.h"
+
+#include "libpspp/str.h"
+#include "data/casereader.h"
+#include "data/dictionary.h"
+#include "data/format.h"
+#include "data/variable.h"
+#include "math/moments.h"
+#include "math/levene.h"
+#include "output/pivot-table.h"
+
+#include "gettext.h"
+#define N_(msgid) msgid
+#define _(msgid) gettext (msgid)
+
+
+struct indep_samples
+{
+ const struct variable *gvar;
+ bool cut;
+ const union value *gval0;
+ const union value *gval1;
+};
+
+struct pair_stats
+{
+ struct moments *mom[2];
+ double lev ;
+ struct levene *nl;
+};
+
+
+static void indep_summary (const struct tt *tt, struct indep_samples *is, const struct pair_stats *ps);
+static void indep_test (const struct tt *tt, const struct pair_stats *ps);
+
+static int
+which_group (const union value *v, const struct indep_samples *is)
+{
+ int width = var_get_width (is->gvar);
+ int cmp = value_compare_3way (v, is->gval0, width);
+ if (is->cut)
+ return (cmp < 0);
+
+ if (cmp == 0)
+ return 0;
+
+ if (0 == value_compare_3way (v, is->gval1, width))
+ return 1;
+
+ return -1;
+}
+
+void
+indep_run (struct tt *tt, const struct variable *gvar,
+ bool cut,
+ const union value *gval0, const union value *gval1,
+ struct casereader *reader)
+{
+ struct indep_samples is;
+ struct ccase *c;
+ struct casereader *r;
+
+ struct pair_stats *ps = XCALLOC (tt->n_vars, struct pair_stats);
+
+ int v;
+
+ for (v = 0; v < tt->n_vars; ++v)
+ {
+ ps[v].mom[0] = moments_create (MOMENT_VARIANCE);
+ ps[v].mom[1] = moments_create (MOMENT_VARIANCE);
+ ps[v].nl = levene_create (var_get_width (gvar), cut ? gval0: NULL);
+ }
+
+ is.gvar = gvar;
+ is.gval0 = gval0;
+ is.gval1 = gval1;
+ is.cut = cut;
+
+ r = casereader_clone (reader);
+ for (; (c = casereader_read (r)); case_unref (c))
+ {
+ double w = dict_get_case_weight (tt->dict, c, NULL);
+
+ const union value *gv = case_data (c, gvar);
+
+ int grp = which_group (gv, &is);
+ if (grp < 0)
+ continue;
+
+ for (v = 0; v < tt->n_vars; ++v)
+ {
+ const union value *val = case_data (c, tt->vars[v]);
+ if (var_is_value_missing (tt->vars[v], val) & tt->exclude)
+ continue;
+
+ moments_pass_one (ps[v].mom[grp], val->f, w);
+ levene_pass_one (ps[v].nl, val->f, w, gv);
+ }
+ }
+ casereader_destroy (r);
+
+ r = casereader_clone (reader);
+ for (; (c = casereader_read (r)); case_unref (c))
+ {
+ double w = dict_get_case_weight (tt->dict, c, NULL);
+
+ const union value *gv = case_data (c, gvar);
+
+ int grp = which_group (gv, &is);
+ if (grp < 0)
+ continue;
+
+ for (v = 0; v < tt->n_vars; ++v)
+ {
+ const union value *val = case_data (c, tt->vars[v]);
+ if (var_is_value_missing (tt->vars[v], val) & tt->exclude)
+ continue;
+
+ moments_pass_two (ps[v].mom[grp], val->f, w);
+ levene_pass_two (ps[v].nl, val->f, w, gv);
+ }
+ }
+ casereader_destroy (r);
+
+ r = reader;
+ for (; (c = casereader_read (r)); case_unref (c))
+ {
+ double w = dict_get_case_weight (tt->dict, c, NULL);
+
+ const union value *gv = case_data (c, gvar);
+
+ int grp = which_group (gv, &is);
+ if (grp < 0)
+ continue;
+
+ for (v = 0; v < tt->n_vars; ++v)
+ {
+ const union value *val = case_data (c, tt->vars[v]);
+ if (var_is_value_missing (tt->vars[v], val) & tt->exclude)
+ continue;
+
+ levene_pass_three (ps[v].nl, val->f, w, gv);
+ }
+ }
+ casereader_destroy (r);
+
+
+ for (v = 0; v < tt->n_vars; ++v)
+ ps[v].lev = levene_calculate (ps[v].nl);
+
+ indep_summary (tt, &is, ps);
+ indep_test (tt, ps);
+
+
+ for (v = 0; v < tt->n_vars; ++v)
+ {
+ moments_destroy (ps[v].mom[0]);
+ moments_destroy (ps[v].mom[1]);
+ levene_destroy (ps[v].nl);
+ }
+ free (ps);
+}
+
+
+static void
+indep_summary (const struct tt *tt, struct indep_samples *is, const struct pair_stats *ps)
+{
+ struct pivot_table *table = pivot_table_create (N_("Group Statistics"));
+ pivot_table_set_weight_var (table, tt->wv);
+
+ pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Statistics"),
+ N_("N"), PIVOT_RC_COUNT,
+ N_("Mean"), PIVOT_RC_OTHER,
+ N_("Std. Deviation"), PIVOT_RC_OTHER,
+ N_("S.E. Mean"), PIVOT_RC_OTHER);
+
+ struct pivot_dimension *group = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Group"));
+ group->root->show_label = true;
+ if (is->cut)
+ {
+ struct string vallab0 = DS_EMPTY_INITIALIZER;
+ ds_put_cstr (&vallab0, "≥");
+ var_append_value_name (is->gvar, is->gval0, &vallab0);
+ pivot_category_create_leaf (group->root,
+ pivot_value_new_user_text_nocopy (
+ ds_steal_cstr (&vallab0)));
+
+ struct string vallab1 = DS_EMPTY_INITIALIZER;
+ ds_put_cstr (&vallab1, "<");
+ var_append_value_name (is->gvar, is->gval0, &vallab1);
+ pivot_category_create_leaf (group->root,
+ pivot_value_new_user_text_nocopy (
+ ds_steal_cstr (&vallab1)));
+ }
+ else
+ {
+ pivot_category_create_leaf (
+ group->root, pivot_value_new_var_value (is->gvar, is->gval0));
+ pivot_category_create_leaf (
+ group->root, pivot_value_new_var_value (is->gvar, is->gval1));
+ }
+
+ struct pivot_dimension *dep_vars = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Dependent Variables"));
+
+ for (size_t v = 0; v < tt->n_vars; ++v)
+ {
+ const struct variable *var = tt->vars[v];
+
+ int dep_var_idx = pivot_category_create_leaf (
+ dep_vars->root, pivot_value_new_variable (var));
+
+ for (int i = 0 ; i < 2; ++i)
+ {
+ double cc, mean, sigma;
+ moments_calculate (ps[v].mom[i], &cc, &mean, &sigma, NULL, NULL);
+
+ double entries[] = { cc, mean, sqrt (sigma), sqrt (sigma / cc) };
+ for (size_t j = 0; j < sizeof entries / sizeof *entries; j++)
+ pivot_table_put3 (table, j, i, dep_var_idx,
+ pivot_value_new_number (entries[j]));
+ }
+ }
+
+ pivot_table_submit (table);
+}
+
+
+static void
+indep_test (const struct tt *tt, const struct pair_stats *ps)
+{
+ struct pivot_table *table = pivot_table_create (
+ N_("Independent Samples Test"));
+
+ struct pivot_dimension *statistics = pivot_dimension_create (
+ table, PIVOT_AXIS_COLUMN, N_("Statistics"));
+ pivot_category_create_group (
+ statistics->root, N_("Levene's Test for Equality of Variances"),
+ N_("F"), PIVOT_RC_OTHER,
+ N_("Sig."), PIVOT_RC_SIGNIFICANCE);
+ struct pivot_category *group = pivot_category_create_group (
+ statistics->root, N_("T-Test for Equality of Means"),
+ N_("t"), PIVOT_RC_OTHER,
+ N_("df"), PIVOT_RC_OTHER,
+ N_("Sig. (2-tailed)"), PIVOT_RC_SIGNIFICANCE,
+ N_("Mean Difference"), PIVOT_RC_OTHER,
+ N_("Std. Error Difference"), PIVOT_RC_OTHER);
+ pivot_category_create_group (
+ /* xgettext:no-c-format */
+ group, N_("95% Confidence Interval of the Difference"),
+ N_("Lower"), PIVOT_RC_OTHER,
+ N_("Upper"), PIVOT_RC_OTHER);
+
+ pivot_dimension_create (table, PIVOT_AXIS_ROW, N_("Assumptions"),
+ N_("Equal variances assumed"),
+ N_("Equal variances not assumed"));
+
+ struct pivot_dimension *dep_vars = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Dependent Variables"));
+
+ for (size_t v = 0; v < tt->n_vars; ++v)
+ {
+ int dep_var_idx = pivot_category_create_leaf (
+ dep_vars->root, pivot_value_new_variable (tt->vars[v]));
+
+ double cc0, mean0, sigma0;
+ double cc1, mean1, sigma1;
+ moments_calculate (ps[v].mom[0], &cc0, &mean0, &sigma0, NULL, NULL);
+ moments_calculate (ps[v].mom[1], &cc1, &mean1, &sigma1, NULL, NULL);
+
+ double mean_diff = mean0 - mean1;
+
+
+ /* Equal variances assumed. */
+ double e_df = cc0 + cc1 - 2.0;
+ double e_pooled_variance = ((cc0 - 1)* sigma0 + (cc1 - 1) * sigma1) / e_df;
+ double e_tval = ((mean0 - mean1) / sqrt (e_pooled_variance)
+ / sqrt ((cc0 + cc1) / (cc0 * cc1)));
+ double e_p = gsl_cdf_tdist_P (e_tval, e_df);
+ double e_q = gsl_cdf_tdist_Q (e_tval, e_df);
+ double e_sig = 2.0 * (e_tval > 0 ? e_q : e_p);
+ double e_std_err_diff = sqrt (e_pooled_variance * (1.0/cc0 + 1.0/cc1));
+ double e_tval_qinv = gsl_cdf_tdist_Qinv ((1 - tt->confidence) / 2.0, e_df);
+
+ /* Equal variances not assumed */
+ const double s0 = sigma0 / cc0;
+ const double s1 = sigma1 / cc1;
+ double d_df = (pow2 (s0 + s1) / (pow2 (s0) / (cc0 - 1)
+ + pow2 (s1) / (cc1 - 1)));
+ double d_tval = mean_diff / sqrt (sigma0 / cc0 + sigma1 / cc1);
+ double d_p = gsl_cdf_tdist_P (d_tval, d_df);
+ double d_q = gsl_cdf_tdist_Q (d_tval, d_df);
+ double d_sig = 2.0 * (d_tval > 0 ? d_q : d_p);
+ double d_std_err_diff = sqrt ((sigma0 / cc0) + (sigma1 / cc1));
+ double d_tval_qinv = gsl_cdf_tdist_Qinv ((1 - tt->confidence) / 2.0, d_df);
+
+ struct entry
+ {
+ int assumption_idx;
+ int stat_idx;
+ double x;
+ }
+ entries[] =
+ {
+ { 0, 0, ps[v].lev },
+ { 0, 1, gsl_cdf_fdist_Q (ps[v].lev, 1, cc0 + cc1 - 2) },
+
+ { 0, 2, e_tval },
+ { 0, 3, e_df },
+ { 0, 4, e_sig },
+ { 0, 5, mean_diff },
+ { 0, 6, e_std_err_diff },
+ { 0, 7, mean_diff - e_tval_qinv * e_std_err_diff },
+ { 0, 8, mean_diff + e_tval_qinv * e_std_err_diff },
+
+ { 1, 2, d_tval },
+ { 1, 3, d_df },
+ { 1, 4, d_sig },
+ { 1, 5, mean_diff },
+ { 1, 6, d_std_err_diff },
+ { 1, 7, mean_diff - d_tval_qinv * d_std_err_diff },
+ { 1, 8, mean_diff + d_tval_qinv * d_std_err_diff },
+ };
+
+ 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->assumption_idx,
+ dep_var_idx, pivot_value_new_number (e->x));
+ }
+ }
+
+ pivot_table_submit (table);
+}
projects / pspp / blobdiff