1 /* PSPP - a program for statistical analysis. -*-c-*-
2 Copyright (C) 2006, 2008, 2009 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/>. */
19 #include <language/stats/npar.h>
23 #include <data/case.h>
24 #include <data/casegrouper.h>
25 #include <data/casereader.h>
26 #include <data/dictionary.h>
27 #include <data/procedure.h>
28 #include <language/command.h>
29 #include <language/lexer/lexer.h>
30 #include <language/lexer/variable-parser.h>
31 #include <language/stats/binomial.h>
32 #include <language/stats/chisquare.h>
33 #include <language/stats/wilcoxon.h>
34 #include <language/stats/sign.h>
35 #include <libpspp/hash.h>
36 #include <libpspp/pool.h>
37 #include <libpspp/taint.h>
38 #include <math/moments.h>
40 #include "npar-summary.h"
43 #define _(msgid) gettext (msgid)
57 missing=miss:!analysis/listwise,
58 incl:include/!exclude;
60 +statistics[st_]=descriptives,quartiles,all.
66 static struct cmd_npar_tests cmd;
72 struct npar_test **test;
75 const struct variable ** vv; /* Compendium of all variables
76 (those mentioned on ANY subcommand */
77 int n_vars; /* Number of variables in vv */
79 enum mv_class filter; /* Missing values to filter. */
81 bool descriptives; /* Descriptive statistics should be calculated */
82 bool quartiles; /* Quartiles should be calculated */
84 bool exact; /* Whether exact calculations have been requested */
85 double timer; /* Maximum time (in minutes) to wait for exact calculations */
88 static void one_sample_insert_variables (const struct npar_test *test,
89 struct const_hsh_table *variables);
91 static void two_sample_insert_variables (const struct npar_test *test,
92 struct const_hsh_table *variables);
97 npar_execute(struct casereader *input,
98 const struct npar_specs *specs,
99 const struct dataset *ds)
102 struct descriptives *summary_descriptives = NULL;
104 for ( t = 0 ; t < specs->n_tests; ++t )
106 const struct npar_test *test = specs->test[t];
107 if ( NULL == test->execute )
109 msg (SW, _("NPAR subcommand not currently implemented."));
112 test->execute (ds, casereader_clone (input), specs->filter, test, specs->exact, specs->timer);
115 if ( specs->descriptives )
117 summary_descriptives = xnmalloc (sizeof (*summary_descriptives),
120 npar_summary_calc_descriptives (summary_descriptives,
121 casereader_clone (input),
123 specs->vv, specs->n_vars,
127 if ( (specs->descriptives || specs->quartiles)
128 && !taint_has_tainted_successor (casereader_get_taint (input)) )
129 do_summary_box (summary_descriptives, specs->vv, specs->n_vars );
131 free (summary_descriptives);
132 casereader_destroy (input);
136 cmd_npar_tests (struct lexer *lexer, struct dataset *ds)
140 struct npar_specs npar_specs = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
141 struct const_hsh_table *var_hash;
142 struct casegrouper *grouper;
143 struct casereader *input, *group;
145 npar_specs.pool = pool_create ();
147 var_hash = const_hsh_create_pool (npar_specs.pool, 0,
148 compare_vars_by_name, hash_var_by_name,
151 if ( ! parse_npar_tests (lexer, ds, &cmd, &npar_specs) )
153 pool_destroy (npar_specs.pool);
157 for (i = 0; i < npar_specs.n_tests; ++i )
159 const struct npar_test *test = npar_specs.test[i];
160 test->insert_variables (test, var_hash);
163 npar_specs.vv = (const struct variable **) const_hsh_sort (var_hash);
164 npar_specs.n_vars = const_hsh_count (var_hash);
166 if ( cmd.sbc_statistics )
170 for ( i = 0 ; i < NPAR_ST_count; ++i )
172 if ( cmd.a_statistics[i] )
176 case NPAR_ST_DESCRIPTIVES:
177 npar_specs.descriptives = true;
179 case NPAR_ST_QUARTILES:
180 npar_specs.quartiles = true;
183 npar_specs.quartiles = true;
184 npar_specs.descriptives = true;
193 npar_specs.filter = cmd.incl == NPAR_EXCLUDE ? MV_ANY : MV_SYSTEM;
195 input = proc_open (ds);
196 if ( cmd.miss == NPAR_LISTWISE )
198 input = casereader_create_filter_missing (input,
206 grouper = casegrouper_create_splits (input, dataset_dict (ds));
207 while (casegrouper_get_next_group (grouper, &group))
208 npar_execute (group, &npar_specs, ds);
209 ok = casegrouper_destroy (grouper);
210 ok = proc_commit (ds) && ok;
212 const_hsh_destroy (var_hash);
214 pool_destroy (npar_specs.pool);
216 return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
220 npar_custom_chisquare (struct lexer *lexer, struct dataset *ds,
221 struct cmd_npar_tests *cmd UNUSED, void *aux )
223 struct npar_specs *specs = aux;
225 struct chisquare_test *cstp = pool_alloc(specs->pool, sizeof(*cstp));
226 struct one_sample_test *tp = &cstp->parent;
227 struct npar_test *nt = &tp->parent;
229 nt->execute = chisquare_execute;
230 nt->insert_variables = one_sample_insert_variables;
232 if (!parse_variables_const_pool (lexer, specs->pool, dataset_dict (ds),
233 &tp->vars, &tp->n_vars,
234 PV_NO_SCRATCH | PV_NO_DUPLICATE))
239 cstp->ranged = false;
241 if ( lex_match (lexer, '('))
244 if ( ! lex_force_num (lexer)) return 0;
245 cstp->lo = lex_integer (lexer);
247 lex_force_match (lexer, ',');
248 if (! lex_force_num (lexer) ) return 0;
249 cstp->hi = lex_integer (lexer);
250 if ( cstp->lo >= cstp->hi )
253 _("The specified value of HI (%d) is "
254 "lower than the specified value of LO (%d)"),
259 if (! lex_force_match (lexer, ')')) return 0;
262 cstp->n_expected = 0;
263 cstp->expected = NULL;
264 if ( lex_match (lexer, '/') )
266 if ( lex_match_id (lexer, "EXPECTED") )
268 lex_force_match (lexer, '=');
269 if ( ! lex_match_id (lexer, "EQUAL") )
273 while ( lex_is_number(lexer) )
277 f = lex_number (lexer);
279 if ( lex_match (lexer, '*'))
282 f = lex_number (lexer);
285 lex_match (lexer, ',');
287 cstp->n_expected += n;
288 cstp->expected = pool_realloc (specs->pool,
292 for ( i = cstp->n_expected - n ;
293 i < cstp->n_expected;
295 cstp->expected[i] = f;
301 lex_put_back (lexer, '/');
304 if ( cstp->ranged && cstp->n_expected > 0 &&
305 cstp->n_expected != cstp->hi - cstp->lo + 1 )
308 _("%d expected values were given, but the specified "
309 "range (%d-%d) requires exactly %d values."),
310 cstp->n_expected, cstp->lo, cstp->hi,
311 cstp->hi - cstp->lo +1);
316 specs->test = pool_realloc (specs->pool,
318 sizeof(*specs->test) * specs->n_tests);
320 specs->test[specs->n_tests - 1] = nt;
327 npar_custom_binomial (struct lexer *lexer, struct dataset *ds,
328 struct cmd_npar_tests *cmd UNUSED, void *aux)
330 struct npar_specs *specs = aux;
331 struct binomial_test *btp = pool_alloc(specs->pool, sizeof(*btp));
332 struct one_sample_test *tp = &btp->parent;
333 struct npar_test *nt = &tp->parent;
335 nt->execute = binomial_execute;
336 nt->insert_variables = one_sample_insert_variables;
338 btp->category1 = btp->category2 = btp->cutpoint = SYSMIS;
340 if ( lex_match(lexer, '(') )
342 if ( lex_force_num (lexer) )
344 btp->p = lex_number (lexer);
346 lex_force_match (lexer, ')');
352 if ( lex_match (lexer, '=') )
354 if (parse_variables_const_pool (lexer, specs->pool, dataset_dict (ds),
355 &tp->vars, &tp->n_vars,
356 PV_NUMERIC | PV_NO_SCRATCH | PV_NO_DUPLICATE) )
358 if ( lex_match (lexer, '('))
360 lex_force_num (lexer);
361 btp->category1 = lex_number (lexer);
363 if ( lex_match (lexer, ','))
365 if ( ! lex_force_num (lexer) ) return 2;
366 btp->category2 = lex_number (lexer);
371 btp->cutpoint = btp->category1;
374 lex_force_match (lexer, ')');
382 specs->test = pool_realloc (specs->pool,
384 sizeof(*specs->test) * specs->n_tests);
386 specs->test[specs->n_tests - 1] = nt;
392 bool parse_two_sample_related_test (struct lexer *lexer,
393 const struct dictionary *dict,
394 struct cmd_npar_tests *cmd,
395 struct two_sample_test *test_parameters,
401 parse_two_sample_related_test (struct lexer *lexer,
402 const struct dictionary *dict,
403 struct cmd_npar_tests *cmd UNUSED,
404 struct two_sample_test *test_parameters,
411 const struct variable **vlist1;
414 const struct variable **vlist2;
417 test_parameters->parent.insert_variables = two_sample_insert_variables;
419 if (!parse_variables_const_pool (lexer, pool,
422 PV_NUMERIC | PV_NO_SCRATCH | PV_NO_DUPLICATE) )
425 if ( lex_match(lexer, T_WITH))
428 if ( !parse_variables_const_pool (lexer, pool, dict,
430 PV_NUMERIC | PV_NO_SCRATCH | PV_NO_DUPLICATE) )
433 paired = (lex_match (lexer, '(') &&
434 lex_match_id (lexer, "PAIRED") && lex_match (lexer, ')'));
442 if ( n_vlist1 != n_vlist2)
443 msg (SE, _("PAIRED was specified but the number of variables "
444 "preceding WITH (%zu) did not match the number "
445 "following (%zu)."), n_vlist1, n_vlist2);
447 test_parameters->n_pairs = n_vlist1 ;
451 test_parameters->n_pairs = n_vlist1 * n_vlist2;
456 test_parameters->n_pairs = (n_vlist1 * (n_vlist1 - 1)) / 2 ;
459 test_parameters->pairs =
460 pool_alloc (pool, sizeof ( variable_pair) * test_parameters->n_pairs);
467 assert (n_vlist1 == n_vlist2);
468 for ( i = 0 ; i < n_vlist1; ++i )
470 test_parameters->pairs[n][1] = vlist1[i];
471 test_parameters->pairs[n][0] = vlist2[i];
478 for ( i = 0 ; i < n_vlist1; ++i )
480 for ( j = 0 ; j < n_vlist2; ++j )
482 test_parameters->pairs[n][1] = vlist1[i];
483 test_parameters->pairs[n][0] = vlist2[j];
492 for ( i = 0 ; i < n_vlist1 - 1; ++i )
494 for ( j = i + 1 ; j < n_vlist1; ++j )
496 assert ( n < test_parameters->n_pairs);
497 test_parameters->pairs[n][1] = vlist1[i];
498 test_parameters->pairs[n][0] = vlist1[j];
504 assert ( n == test_parameters->n_pairs);
510 npar_custom_wilcoxon (struct lexer *lexer,
512 struct cmd_npar_tests *cmd, void *aux )
514 struct npar_specs *specs = aux;
516 struct two_sample_test *tp = pool_alloc (specs->pool, sizeof(*tp));
517 struct npar_test *nt = &tp->parent;
518 nt->execute = wilcoxon_execute;
520 if (!parse_two_sample_related_test (lexer, dataset_dict (ds), cmd,
525 specs->test = pool_realloc (specs->pool,
527 sizeof(*specs->test) * specs->n_tests);
528 specs->test[specs->n_tests - 1] = nt;
534 npar_custom_mcnemar (struct lexer *lexer,
536 struct cmd_npar_tests *cmd, void *aux )
538 struct npar_specs *specs = aux;
540 struct two_sample_test *tp = pool_alloc(specs->pool, sizeof(*tp));
541 struct npar_test *nt = &tp->parent;
545 if (!parse_two_sample_related_test (lexer, dataset_dict (ds),
546 cmd, tp, specs->pool) )
550 specs->test = pool_realloc (specs->pool,
552 sizeof(*specs->test) * specs->n_tests);
553 specs->test[specs->n_tests - 1] = nt;
559 npar_custom_sign (struct lexer *lexer, struct dataset *ds,
560 struct cmd_npar_tests *cmd, void *aux )
562 struct npar_specs *specs = aux;
564 struct two_sample_test *tp = pool_alloc(specs->pool, sizeof(*tp));
565 struct npar_test *nt = &tp->parent;
567 nt->execute = sign_execute;
569 if (!parse_two_sample_related_test (lexer, dataset_dict (ds), cmd,
574 specs->test = pool_realloc (specs->pool,
576 sizeof(*specs->test) * specs->n_tests);
577 specs->test[specs->n_tests - 1] = nt;
582 /* Insert the variables for TEST into VAR_HASH */
584 one_sample_insert_variables (const struct npar_test *test,
585 struct const_hsh_table *var_hash)
588 struct one_sample_test *ost = (struct one_sample_test *) test;
590 for ( i = 0 ; i < ost->n_vars ; ++i )
591 const_hsh_insert (var_hash, ost->vars[i]);
595 two_sample_insert_variables (const struct npar_test *test,
596 struct const_hsh_table *var_hash)
600 const struct two_sample_test *tst = (const struct two_sample_test *) test;
602 for ( i = 0 ; i < tst->n_pairs ; ++i )
604 variable_pair *pair = &tst->pairs[i];
606 const_hsh_insert (var_hash, (*pair)[0]);
607 const_hsh_insert (var_hash, (*pair)[1]);
614 npar_custom_method (struct lexer *lexer, struct dataset *ds UNUSED,
615 struct cmd_npar_tests *test UNUSED, void *aux)
617 struct npar_specs *specs = aux;
619 if ( lex_match_id (lexer, "EXACT") )
623 if (lex_match_id (lexer, "TIMER"))
627 if ( lex_match (lexer, '('))
629 if ( lex_force_num (lexer) )
631 specs->timer = lex_number (lexer);
634 lex_force_match (lexer, ')');