1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 2011 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>.
21 #include <gsl/gsl_cdf.h>
23 #include "data/format.h"
26 #include "data/variable.h"
27 #include "data/case.h"
28 #include "data/dictionary.h"
29 #include "data/dataset.h"
30 #include "data/casereader.h"
31 #include "data/casewriter.h"
32 #include "data/subcase.h"
33 #include "data/value.h"
35 #include "math/percentiles.h"
36 #include "math/sort.h"
38 #include "libpspp/cast.h"
39 #include "libpspp/hmap.h"
40 #include "libpspp/array.h"
41 #include "libpspp/str.h"
42 #include "libpspp/misc.h"
44 #include "output/tab.h"
47 #define _(msgid) gettext (msgid)
52 struct hmap_node node;
60 const struct variable *var;
61 struct val_node **sorted_array;
70 val_node_cmp_3way (const void *a_, const void *b_, const void *aux)
72 const struct variable *indep_var = aux;
73 const struct val_node *const *a = a_;
74 const struct val_node *const *b = b_;
76 return value_compare_3way (&(*a)->val, &(*b)->val, var_get_width (indep_var));
80 show_frequencies (const struct n_sample_test *nst, const struct results *results, int n_vals, const struct dictionary *);
83 show_test_statistics (const struct n_sample_test *nst, const struct results *results, int, const struct dictionary *);
86 static struct val_node *
87 find_value (const struct hmap *map, const union value *val,
88 const struct variable *var)
90 struct val_node *foo = NULL;
91 size_t hash = value_hash (val, var_get_width (var), 0);
92 HMAP_FOR_EACH_WITH_HASH (foo, struct val_node, node, hash, map)
93 if (value_equal (val, &foo->val, var_get_width (var)))
100 median_execute (const struct dataset *ds,
101 struct casereader *input,
102 enum mv_class exclude,
103 const struct npar_test *test,
107 const struct dictionary *dict = dataset_dict (ds);
108 const struct variable *wvar = dict_get_weight (dict);
111 const struct median_test *mt = UP_CAST (test, const struct median_test,
114 const struct n_sample_test *nst = UP_CAST (test, const struct n_sample_test,
117 const bool n_sample_test = (value_compare_3way (&nst->val2, &nst->val1,
118 var_get_width (nst->indep_var)) > 0);
120 struct results *results = xcalloc (nst->n_vars, sizeof (*results));
122 for (v = 0; v < nst->n_vars; ++v)
126 double median = mt->median;
127 const struct variable *var = nst->vars[v];
129 struct hmap map = HMAP_INITIALIZER (map);
130 struct casereader *r = casereader_clone (input);
134 if (n_sample_test == false)
136 struct val_node *vn = xzalloc (sizeof *vn);
137 value_clone (&vn->val, &nst->val1, var_get_width (nst->indep_var));
138 hmap_insert (&map, &vn->node, value_hash (&nst->val1,
139 var_get_width (nst->indep_var), 0));
141 vn = xzalloc (sizeof *vn);
142 value_clone (&vn->val, &nst->val2, var_get_width (nst->indep_var));
143 hmap_insert (&map, &vn->node, value_hash (&nst->val2,
144 var_get_width (nst->indep_var), 0));
147 if ( median == SYSMIS)
149 struct percentile *ptl;
150 struct order_stats *os;
152 struct casereader *rr;
154 struct casewriter *writer;
155 subcase_init_var (&sc, var, SC_ASCEND);
156 rr = casereader_clone (r);
157 writer = sort_create_writer (&sc, casereader_get_proto (rr));
159 for (; (c = casereader_read (rr)) != NULL; )
161 if ( var_is_value_missing (var, case_data (c, var), exclude))
167 cc += dict_get_case_weight (dict, c, &warn);
168 casewriter_write (writer, c);
170 subcase_destroy (&sc);
171 casereader_destroy (rr);
173 rr = casewriter_make_reader (writer);
175 ptl = percentile_create (0.5, cc);
178 order_stats_accumulate (&os, 1,
184 median = percentile_calculate (ptl, PC_HAVERAGE);
185 statistic_destroy (&ptl->parent.parent);
188 results[v].median = median;
191 for (; (c = casereader_read (r)) != NULL; case_unref (c))
193 struct val_node *vn ;
194 const double weight = dict_get_case_weight (dict, c, &warn);
195 const union value *val = case_data (c, var);
196 const union value *indep_val = case_data (c, nst->indep_var);
198 if ( var_is_value_missing (var, case_data (c, var), exclude))
205 int width = var_get_width (nst->indep_var);
206 /* Ignore out of range values */
208 value_compare_3way (indep_val, &nst->val1, width) < 0
210 value_compare_3way (indep_val, &nst->val2, width) > 0
217 vn = find_value (&map, indep_val, nst->indep_var);
220 if ( n_sample_test == true)
222 int width = var_get_width (nst->indep_var);
223 vn = xzalloc (sizeof *vn);
224 value_clone (&vn->val, indep_val, width);
226 hmap_insert (&map, &vn->node, value_hash (indep_val, width, 0));
234 if (val->f <= median)
241 casereader_destroy (r);
245 struct val_node *vn = NULL;
248 HMAP_FOR_EACH (vn, struct val_node, node, &map)
254 results[v].n = count;
255 results[v].sorted_array = xcalloc (hmap_count (&map), sizeof (void*));
256 results[v].var = var;
258 HMAP_FOR_EACH (vn, struct val_node, node, &map)
260 double e_0j = r_0 * (vn->le + vn->gt) / count;
261 double e_1j = r_1 * (vn->le + vn->gt) / count;
263 results[v].chisq += pow2 (vn->le - e_0j) / e_0j;
264 results[v].chisq += pow2 (vn->gt - e_1j) / e_1j;
266 results[v].sorted_array[x++] = vn;
272 sort (results[v].sorted_array, x, sizeof (*results[v].sorted_array),
273 val_node_cmp_3way, nst->indep_var);
278 casereader_destroy (input);
280 show_frequencies (nst, results, n_vals, dict);
281 show_test_statistics (nst, results, n_vals, dict);
283 for (v = 0; v < nst->n_vars; ++v)
286 const struct results *rs = results + v;
288 for (i = 0; i < n_vals; ++i)
290 struct val_node *vn = rs->sorted_array[i];
291 value_destroy (&vn->val, var_get_width (nst->indep_var));
294 free (rs->sorted_array);
302 show_frequencies (const struct n_sample_test *nst, const struct results *results, int n_vals, const struct dictionary *dict)
304 const struct variable *weight = dict_get_weight (dict);
305 const struct fmt_spec *wfmt = weight ? var_get_print_format (weight) : &F_8_0;
310 const int row_headers = 2;
311 const int column_headers = 2;
312 const int nc = row_headers + n_vals;
313 const int nr = column_headers + nst->n_vars * 2;
315 struct tab_table *table = tab_create (nc, nr);
317 tab_headers (table, row_headers, 0, column_headers, 0);
319 tab_title (table, _("Frequencies"));
321 /* Box around the table and vertical lines inside*/
322 tab_box (table, TAL_2, TAL_2, -1, TAL_1,
323 0, 0, tab_nc (table) - 1, tab_nr (table) - 1 );
325 tab_hline (table, TAL_2, 0, tab_nc (table) -1, column_headers);
326 tab_vline (table, TAL_2, row_headers, 0, tab_nr (table) - 1);
328 tab_joint_text (table,
329 row_headers, 0, row_headers + n_vals - 1, 0,
330 TAT_TITLE | TAB_CENTER, var_to_string (nst->indep_var));
333 tab_hline (table, TAL_1, row_headers, tab_nc (table) - 1, 1);
336 for (i = 0; i < n_vals; ++i)
338 const struct results *rs = results + 0;
340 ds_init_empty (&label);
342 var_append_value_name (nst->indep_var, &rs->sorted_array[i]->val,
345 tab_text (table, row_headers + i, 1,
346 TAT_TITLE | TAB_LEFT, ds_cstr (&label));
351 for (v = 0; v < nst->n_vars; ++v)
353 const struct results *rs = &results[v];
354 tab_text (table, 0, column_headers + v * 2,
355 TAT_TITLE | TAB_LEFT, var_to_string (rs->var) );
357 tab_text (table, 1, column_headers + v * 2,
358 TAT_TITLE | TAB_LEFT, _("> Median") );
360 tab_text (table, 1, column_headers + v * 2 + 1,
361 TAT_TITLE | TAB_LEFT, _("≤ Median") );
364 tab_hline (table, TAL_1, 0, tab_nc (table) - 1, column_headers + v * 2);
367 for (v = 0; v < nst->n_vars; ++v)
370 const struct results *rs = &results[v];
372 for (i = 0; i < n_vals; ++i)
374 const struct val_node *vn = rs->sorted_array[i];
375 tab_double (table, row_headers + i, column_headers + v * 2,
378 tab_double (table, row_headers + i, column_headers + v * 2 + 1,
388 show_test_statistics (const struct n_sample_test *nst,
389 const struct results *results,
391 const struct dictionary *dict)
393 const struct variable *weight = dict_get_weight (dict);
394 const struct fmt_spec *wfmt = weight ? var_get_print_format (weight) : &F_8_0;
398 const int row_headers = 1;
399 const int column_headers = 1;
400 const int nc = row_headers + 5;
401 const int nr = column_headers + nst->n_vars;
403 struct tab_table *table = tab_create (nc, nr);
405 tab_headers (table, row_headers, 0, column_headers, 0);
407 tab_title (table, _("Test Statistics"));
410 tab_box (table, TAL_2, TAL_2, -1, TAL_1,
411 0, 0, tab_nc (table) - 1, tab_nr (table) - 1 );
413 tab_hline (table, TAL_2, 0, tab_nc (table) -1, column_headers);
414 tab_vline (table, TAL_2, row_headers, 0, tab_nr (table) - 1);
416 tab_text (table, row_headers + 0, 0,
417 TAT_TITLE | TAB_CENTER, _("N"));
419 tab_text (table, row_headers + 1, 0,
420 TAT_TITLE | TAB_CENTER, _("Median"));
422 tab_text (table, row_headers + 2, 0,
423 TAT_TITLE | TAB_CENTER, _("Chi-Square"));
425 tab_text (table, row_headers + 3, 0,
426 TAT_TITLE | TAB_CENTER, _("df"));
428 tab_text (table, row_headers + 4, 0,
429 TAT_TITLE | TAB_CENTER, _("Asymp. Sig."));
432 for (v = 0; v < nst->n_vars; ++v)
434 double df = n_vals - 1;
435 const struct results *rs = &results[v];
436 tab_text (table, 0, column_headers + v,
437 TAT_TITLE | TAB_LEFT, var_to_string (rs->var));
440 tab_double (table, row_headers + 0, column_headers + v,
443 tab_double (table, row_headers + 1, column_headers + v,
444 0, rs->median, NULL);
446 tab_double (table, row_headers + 2, column_headers + v,
449 tab_double (table, row_headers + 3, column_headers + v,
452 tab_double (table, row_headers + 4, column_headers + v,
453 0, gsl_cdf_chisq_Q (rs->chisq, df), NULL);