X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Ft-test-paired.c;h=921233b68e52b8364c3c9e14e3496c8046cd88fd;hb=5cab4cf3322f29c0ed7134d23740e07382914f20;hp=4935e9a8cb866bd112d5d6118050af97f016ce46;hpb=37f0de7f0ee1870cf5fd4b802be092be844a5b95;p=pspp

diff --git a/src/language/stats/t-test-paired.c b/src/language/stats/t-test-paired.c
index 4935e9a8cb..921233b68e 100644
--- a/src/language/stats/t-test-paired.c
+++ b/src/language/stats/t-test-paired.c
@@ -27,18 +27,16 @@
 #include "data/dictionary.h"
 #include "data/format.h"
 #include "data/variable.h"
-#include "libpspp/hmapx.h"
-#include "libpspp/hash-functions.h"
 
-#include "output/tab.h"
+#include "output/pivot-table.h"
 
 #include "gettext.h"
+#define N_(msgid) msgid
 #define _(msgid) gettext (msgid)
 
 
 struct pair_stats
 {
-  int posn;
   double sum_of_prod;
   struct moments *mom0;
   const struct variable *var0;
@@ -51,7 +49,8 @@ struct pair_stats
 
 struct paired_samp
 {
-  struct hmapx hmap;
+  struct pair_stats *ps;
+  size_t n_ps;
 };
 
 static void paired_summary (const struct tt *tt, struct paired_samp *os);
@@ -61,31 +60,21 @@ static void paired_test (const struct tt *tt, const struct paired_samp *os);
 void
 paired_run (const struct tt *tt, size_t n_pairs, vp *pairs, struct casereader *reader)
 {
-  int i;
   struct ccase *c;
   struct paired_samp ps;
   struct casereader *r;
-  struct hmapx_node *node;
-  struct pair_stats *pp = NULL;
 
-  hmapx_init (&ps.hmap);
-
-  for (i = 0; i < n_pairs; ++i)
+  ps.ps = xcalloc (n_pairs, sizeof *ps.ps);
+  ps.n_ps = n_pairs;
+  for (size_t i = 0; i < n_pairs; ++i)
     {
       vp *pair = &pairs[i];
-      unsigned int hash;
-      struct pair_stats *pp = xzalloc (sizeof *pp);
-      pp->posn = i;
+      struct pair_stats *pp = &ps.ps[i];
       pp->var0 = (*pair)[0];
       pp->var1 = (*pair)[1];
       pp->mom0 = moments_create (MOMENT_VARIANCE);
       pp->mom1 = moments_create (MOMENT_VARIANCE);
       pp->mom_diff = moments_create (MOMENT_VARIANCE);
-
-      hash = hash_pointer ((*pair)[0], 0);
-      hash = hash_pointer ((*pair)[1], hash);
-
-      hmapx_insert (&ps.hmap, pp, hash);
     }
 
   r = casereader_clone (reader);
@@ -93,10 +82,9 @@ paired_run (const struct tt *tt, size_t n_pairs, vp *pairs, struct casereader *r
     {
       double w = dict_get_case_weight (tt->dict, c, NULL);
 
-      struct hmapx_node *node;
-      struct pair_stats *pp = NULL;
-      HMAPX_FOR_EACH (pp, node, &ps.hmap)
+      for (int i = 0; i < ps.n_ps; i++)
 	{
+          struct pair_stats *pp = &ps.ps[i];
 	  const union value *val0 = case_data (c, pp->var0);
 	  const union value *val1 = case_data (c, pp->var1);
           if (var_is_value_missing (pp->var0, val0, tt->exclude))
@@ -117,10 +105,9 @@ paired_run (const struct tt *tt, size_t n_pairs, vp *pairs, struct casereader *r
     {
       double w = dict_get_case_weight (tt->dict, c, NULL);
 
-      struct hmapx_node *node;
-      struct pair_stats *pp = NULL;
-      HMAPX_FOR_EACH (pp, node, &ps.hmap)
+      for (int i = 0; i < ps.n_ps; i++)
 	{
+          struct pair_stats *pp = &ps.ps[i];
 	  const union value *val0 = case_data (c, pp->var0);
 	  const union value *val1 = case_data (c, pp->var1);
           if (var_is_value_missing (pp->var0, val0, tt->exclude))
@@ -142,214 +129,172 @@ paired_run (const struct tt *tt, size_t n_pairs, vp *pairs, struct casereader *r
   paired_test (tt, &ps);
 
   /* Clean up */
-  HMAPX_FOR_EACH (pp, node, &ps.hmap)
+
+  for (int i = 0; i < ps.n_ps; i++)
     {
+      struct pair_stats *pp = &ps.ps[i];
       moments_destroy (pp->mom0);
       moments_destroy (pp->mom1);
       moments_destroy (pp->mom_diff);
-      free (pp);
     }
-
-  hmapx_destroy (&ps.hmap);
+  free (ps.ps);
 }
 
 static void
 paired_summary (const struct tt *tt, struct paired_samp *os)
 {
-  size_t n_pairs = hmapx_count (&os->hmap);
-  struct hmapx_node *node;
-  struct pair_stats *pp = NULL;
-
-  const int heading_rows = 1;
-  const int heading_cols = 2;
-
-  const int cols = 4 + heading_cols;
-  const int rows = n_pairs * 2 + heading_rows;
-  struct tab_table *t = tab_create (cols, rows);
-  const struct fmt_spec *wfmt = tt->wv ? var_get_print_format (tt->wv) : & F_8_0;
-  tab_set_format (t, RC_WEIGHT, wfmt);
-  tab_headers (t, 0, 0, heading_rows, 0);
-  tab_box (t, TAL_2, TAL_2, TAL_0, TAL_0, 0, 0, cols - 1, rows - 1);
-  tab_box (t, -1, -1,       TAL_0, TAL_1, heading_cols, 0, cols - 1, rows - 1);
-
-  tab_hline (t, TAL_2, 0, cols - 1, 1);
-
-  tab_title (t, _("Paired Sample Statistics"));
-  tab_vline (t, TAL_2, heading_cols, 0, rows - 1);
-  tab_text (t, 3, 0, TAB_CENTER | TAT_TITLE, _("N"));
-  tab_text (t, 2, 0, TAB_CENTER | TAT_TITLE, _("Mean"));
-  tab_text (t, 4, 0, TAB_CENTER | TAT_TITLE, _("Std. Deviation"));
-  tab_text (t, 5, 0, TAB_CENTER | TAT_TITLE, _("S.E. Mean"));
-
-  HMAPX_FOR_EACH (pp, node, &os->hmap)
-    {
-      int v = pp->posn;
-      double cc, mean, sigma;
+  struct pivot_table *table = pivot_table_create (
+    N_("Paired Sample Statistics"));
+  pivot_table_set_weight_var (table, tt->wv);
 
-      tab_text_format (t, 0, v * 2 + heading_rows, TAB_LEFT, _("Pair %d"), pp->posn + 1);
-
-      /* first var */
-      moments_calculate (pp->mom0, &cc, &mean, &sigma, NULL, NULL);
-      tab_text (t, 1, v * 2 + heading_rows, TAB_LEFT, var_to_string (pp->var0));
-      tab_double (t, 3, v * 2 + heading_rows, TAB_RIGHT, cc, NULL, RC_WEIGHT);
-      tab_double (t, 2, v * 2 + heading_rows, TAB_RIGHT, mean, NULL, RC_OTHER);
-      tab_double (t, 4, v * 2 + heading_rows, TAB_RIGHT, sqrt (sigma), NULL, RC_OTHER);
-      tab_double (t, 5, v * 2 + heading_rows, TAB_RIGHT, sqrt (sigma / cc), NULL, RC_OTHER);
-
-      /* second var */
-      moments_calculate (pp->mom1, &cc, &mean, &sigma, NULL, NULL);
-      tab_text (t, 1, v * 2 + 1 + heading_rows, TAB_LEFT, var_to_string (pp->var1));
-      tab_double (t, 3, v * 2 + 1 + heading_rows, TAB_RIGHT, cc, NULL, RC_WEIGHT);
-      tab_double (t, 2, v * 2 + 1 + heading_rows, TAB_RIGHT, mean, NULL, RC_OTHER);
-      tab_double (t, 4, v * 2 + 1 + heading_rows, TAB_RIGHT, sqrt (sigma), NULL, RC_OTHER);
-      tab_double (t, 5, v * 2 + 1 + heading_rows, TAB_RIGHT, sqrt (sigma / cc), NULL, RC_OTHER);
+  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 *variables = pivot_dimension_create (
+    table, PIVOT_AXIS_ROW, N_("Variables"));
+
+  for (size_t i = 0; i < os->n_ps; i++)
+    {
+      struct pair_stats *pp = &os->ps[i];
+      struct pivot_category *pair = pivot_category_create_group__ (
+        variables->root, pivot_value_new_text_format (N_("Pair %d"), i + 1));
+
+      for (int j = 0; j < 2; j++)
+        {
+          const struct variable *var = j ? pp->var1 : pp->var0;
+          const struct moments *mom = j ? pp->mom1 : pp->mom0;
+          double cc, mean, sigma;
+          moments_calculate (mom, &cc, &mean, &sigma, NULL, NULL);
+
+          int var_idx = pivot_category_create_leaf (
+            pair, pivot_value_new_variable (var));
+
+          double entries[] = { cc, mean, sqrt (sigma), sqrt (sigma / cc) };
+          for (size_t j = 0; j < sizeof entries / sizeof *entries; j++)
+            pivot_table_put2 (table, j, var_idx,
+                              pivot_value_new_number (entries[j]));
+        }
     }
 
-  tab_submit (t);
+  pivot_table_submit (table);
 }
 
 
 static void
 paired_correlations (const struct tt *tt, struct paired_samp *os)
 {
-  size_t n_pairs = hmapx_count (&os->hmap);
-  struct hmapx_node *node;
-  struct pair_stats *pp = NULL;
-  const int heading_rows = 1;
-  const int heading_cols = 2;
-
-  const int cols = 5;
-  const int rows = n_pairs + heading_rows;
-  struct tab_table *t = tab_create (cols, rows);
-  const struct fmt_spec *wfmt = tt->wv ? var_get_print_format (tt->wv) : & F_8_0;
-  tab_set_format (t, RC_WEIGHT, wfmt);
-  tab_headers (t, 0, 0, heading_rows, 0);
-  tab_box (t, TAL_2, TAL_2, TAL_0, TAL_1, 0, 0, cols - 1, rows - 1);
-
-  tab_hline (t, TAL_2, 0, cols - 1, 1);
-
-  tab_title (t, _("Paired Samples Correlations"));
-  tab_vline (t, TAL_2, heading_cols, 0, rows - 1);
-  tab_text (t, 2, 0, TAB_CENTER | TAT_TITLE, _("N"));
-  tab_text (t, 3, 0, TAB_CENTER | TAT_TITLE, _("Correlation"));
-  tab_text (t, 4, 0, TAB_CENTER | TAT_TITLE, _("Sig."));
-
-  HMAPX_FOR_EACH (pp, node, &os->hmap)
-    {
-      double corr;
-      double cc0, mean0, sigma0;
-      double cc1, mean1, sigma1;
-      int v = pp->posn;
+  struct pivot_table *table = pivot_table_create (
+    N_("Paired Samples Correlations"));
+  pivot_table_set_weight_var (table, tt->wv);
+
+  pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Statistics"),
+                          N_("N"), PIVOT_RC_COUNT,
+                          N_("Correlation"), PIVOT_RC_CORRELATION,
+                          N_("Sig."), PIVOT_RC_SIGNIFICANCE);
 
-      tab_text_format (t, 0, v + heading_rows, TAB_LEFT, _("Pair %d"), pp->posn + 1);
+  struct pivot_dimension *pairs = pivot_dimension_create (
+    table, PIVOT_AXIS_ROW, N_("Pairs"));
 
-      tab_text_format (t, 1, v + heading_rows, TAB_LEFT, _("%s & %s"),
-		       var_to_string (pp->var0),
-		       var_to_string (pp->var1));
+  for (size_t i = 0; i < os->n_ps; i++)
+    {
+      struct pair_stats *pp = &os->ps[i];
+      struct pivot_category *group = pivot_category_create_group__ (
+        pairs->root, pivot_value_new_text_format (N_("Pair %d"), i + 1));
+
+      int row = pivot_category_create_leaf (
+        group, pivot_value_new_text_format (N_("%s & %s"),
+                                            var_to_string (pp->var0),
+                                            var_to_string (pp->var1)));
 
+      double cc0, mean0, sigma0;
+      double cc1, mean1, sigma1;
       moments_calculate (pp->mom0, &cc0, &mean0, &sigma0, NULL, NULL);
       moments_calculate (pp->mom1, &cc1, &mean1, &sigma1, NULL, NULL);
-
       /* If this fails, then we're not dealing with missing values properly */
       assert (cc0 == cc1);
 
-      tab_double (t, 2, v + heading_rows, TAB_RIGHT, cc0, NULL, RC_WEIGHT);
-
-      corr = pp->sum_of_prod / cc0  - (mean0 * mean1);
-      corr /= sqrt (sigma0 * sigma1);
-      corr *= cc0 / (cc0 - 1);
-
-      tab_double (t, 3, v + heading_rows, TAB_RIGHT, corr, NULL, RC_OTHER);
-      tab_double (t, 4, v + heading_rows, TAB_RIGHT,
-		  2.0 * significance_of_correlation (corr, cc0), NULL, RC_PVALUE);
+      double corr = ((pp->sum_of_prod / cc0 - mean0 * mean1)
+                     / sqrt (sigma0 * sigma1) * cc0 / (cc0 - 1));
+      double sig = 2.0 * significance_of_correlation (corr, cc0);
+      double entries[] = { cc0, corr, sig };
+      for (size_t i = 0; i < sizeof entries / sizeof *entries; i++)
+        pivot_table_put2 (table, i, row, pivot_value_new_number (entries[i]));
     }
 
-  tab_submit (t);
+  pivot_table_submit (table);
 }
 
 
 static void
 paired_test (const struct tt *tt, const struct paired_samp *os)
 {
-  size_t n_pairs = hmapx_count (&os->hmap);
-  struct hmapx_node *node;
-  struct pair_stats *pp = NULL;
-
-  const int heading_rows = 3;
-  const int heading_cols = 2;
-  const size_t rows = heading_rows + n_pairs;
-  const size_t cols = 10;
-  const struct fmt_spec *wfmt = tt->wv ? var_get_print_format (tt->wv) : & F_8_0;
-
-  struct tab_table *t = tab_create (cols, rows);
-  tab_set_format (t, RC_WEIGHT, wfmt);
-  tab_headers (t, 0, 0, heading_rows, 0);
-  tab_box (t, TAL_2, TAL_2, TAL_0, TAL_0, 0, 0, cols - 1, rows - 1);
-  tab_hline (t, TAL_2, 0, cols - 1, 3);
-
-  tab_title (t, _("Paired Samples Test"));
-  tab_hline (t, TAL_1, heading_cols, 6, 1);
-  tab_vline (t, TAL_2, heading_cols, 0, rows - 1);
-
-  tab_box (t, -1, -1, -1, TAL_1, heading_cols, 0, cols - 1, rows - 1);
-
-  tab_joint_text (t, 2, 0, 6, 0, TAB_CENTER,
-		  _("Paired Differences"));
-
-  tab_joint_text_format (t, 5, 1, 6, 1, TAB_CENTER,
-                         _("%g%% Confidence Interval of the Difference"),
-                         tt->confidence * 100.0);
-
-  tab_hline (t, TAL_1, 5, 6, 2);
-  tab_text (t, 7, 2, TAB_CENTER | TAT_TITLE, _("t"));
-  tab_text (t, 8, 2, TAB_CENTER | TAT_TITLE, _("df"));
-  tab_text (t, 9, 2, TAB_CENTER | TAT_TITLE, _("Sig. (2-tailed)"));
-  tab_text (t, 4, 2, TAB_CENTER | TAT_TITLE, _("Std. Error Mean"));
-  tab_text (t, 3, 2, TAB_CENTER | TAT_TITLE, _("Std. Deviation"));
-  tab_text (t, 2, 2, TAB_CENTER | TAT_TITLE, _("Mean"));
-
-  tab_text (t, 5, 2, TAB_CENTER | TAT_TITLE, _("Lower"));
-  tab_text (t, 6, 2, TAB_CENTER | TAT_TITLE, _("Upper"));
-
-  HMAPX_FOR_EACH (pp, node, &os->hmap)
+  struct pivot_table *table = pivot_table_create (N_("Paired Samples Test"));
+  pivot_table_set_weight_var (table, tt->wv);
+
+  struct pivot_dimension *statistics = pivot_dimension_create (
+    table, PIVOT_AXIS_COLUMN, N_("Statistics"));
+  struct pivot_category *group = pivot_category_create_group (
+    statistics->root, N_("Paired Differences"),
+    N_("Mean"), PIVOT_RC_OTHER,
+    N_("Std. Deviation"), PIVOT_RC_OTHER,
+    N_("S.E. Mean"), PIVOT_RC_OTHER);
+  struct pivot_category *interval = pivot_category_create_group__ (
+    group, pivot_value_new_text_format (
+      N_("%g%% Confidence Interval of the Difference"),
+      tt->confidence * 100.0));
+  pivot_category_create_leaves (interval,
+                                N_("Lower"), PIVOT_RC_OTHER,
+                                N_("Upper"), PIVOT_RC_OTHER);
+  pivot_category_create_leaves (statistics->root,
+                                N_("t"), PIVOT_RC_OTHER,
+                                N_("df"), PIVOT_RC_COUNT,
+                                N_("Sig. (2-tailed)"), PIVOT_RC_SIGNIFICANCE);
+
+  struct pivot_dimension *pairs = pivot_dimension_create (
+    table, PIVOT_AXIS_ROW, N_("Pairs"));
+
+  for (size_t i = 0; i < os->n_ps; i++)
     {
-      int v = pp->posn;
-      double cc, mean, sigma;
-      double df ;
-      double tval;
-      double p, q;
-      double se_mean;
-
-      moments_calculate (pp->mom_diff, &cc, &mean, &sigma, NULL, NULL);
-
-      df = cc - 1.0;
-      tab_text_format (t, 0, v + heading_rows, TAB_LEFT, _("Pair %d"), v + 1);
+      struct pair_stats *pp = &os->ps[i];
+      struct pivot_category *group = pivot_category_create_group__ (
+        pairs->root, pivot_value_new_text_format (N_("Pair %d"), i + 1));
 
-      tab_text_format (t, 1, v + heading_rows, TAB_LEFT, _("%s - %s"),
-		       var_to_string (pp->var0),
-		       var_to_string (pp->var1));
+      int row = pivot_category_create_leaf (
+        group, pivot_value_new_text_format (N_("%s - %s"),
+                                            var_to_string (pp->var0),
+                                            var_to_string (pp->var1)));
 
-      tval = mean * sqrt (cc / sigma);
-      se_mean = sqrt (sigma / cc);
-
-      tab_double (t, 2, v + heading_rows, TAB_RIGHT, mean, NULL, RC_OTHER);
-      tab_double (t, 3, v + heading_rows, TAB_RIGHT, sqrt (sigma), NULL, RC_OTHER);
-      tab_double (t, 4, v + heading_rows, TAB_RIGHT, se_mean, NULL, RC_OTHER);
+      double cc, mean, sigma;
+      moments_calculate (pp->mom_diff, &cc, &mean, &sigma, NULL, NULL);
 
-      tab_double (t, 7, v + heading_rows, TAB_RIGHT, tval, NULL, RC_OTHER);
-      tab_double (t, 8, v + heading_rows, TAB_RIGHT, df, NULL, RC_WEIGHT);
+      double df = cc - 1.0;
 
+      double t = mean * sqrt (cc / sigma);
+      double se_mean = sqrt (sigma / cc);
 
-      p = gsl_cdf_tdist_P (tval, df);
-      q = gsl_cdf_tdist_Q (tval, df);
+      double p = gsl_cdf_tdist_P (t, df);
+      double q = gsl_cdf_tdist_Q (t, df);
+      double sig = 2.0 * (t > 0 ? q : p);
 
-      tab_double (t, 9, v + heading_rows, TAB_RIGHT, 2.0 * (tval > 0 ? q : p), NULL, RC_PVALUE);
+      double t_qinv = gsl_cdf_tdist_Qinv ((1.0 - tt->confidence) / 2.0, df);
 
-      tval = gsl_cdf_tdist_Qinv ( (1.0 - tt->confidence) / 2.0, df);
+      double entries[] = {
+        mean,
+        sqrt (sigma),
+        se_mean,
+        mean - t_qinv * se_mean,
+        mean + t_qinv * se_mean,
+        t,
+        df,
+        sig,
+      };
+      for (size_t i = 0; i < sizeof entries / sizeof *entries; i++)
+        pivot_table_put2 (table, i, row, pivot_value_new_number (entries[i]));
 
-      tab_double (t, 5, v + heading_rows, TAB_RIGHT, mean - tval * se_mean, NULL, RC_OTHER);
-      tab_double (t, 6, v + heading_rows, TAB_RIGHT, mean + tval * se_mean, NULL, RC_OTHER);
     }
 
-  tab_submit (t);
+  pivot_table_submit (table);
 }