1 /* PSPP - a program for statistical analysis. -*-c-*-
2 Copyright (C) 2006, 2008 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 <libpspp/hash.h>
35 #include <libpspp/pool.h>
36 #include <libpspp/taint.h>
37 #include <math/moments.h>
39 #include "npar-summary.h"
42 #define _(msgid) gettext (msgid)
56 missing=miss:!analysis/listwise,
57 incl:include/!exclude;
59 +statistics[st_]=descriptives,quartiles,all.
65 static struct cmd_npar_tests cmd;
71 struct npar_test **test;
74 const struct variable ** vv; /* Compendium of all variables
75 (those mentioned on ANY subcommand */
76 int n_vars; /* Number of variables in vv */
78 enum mv_class filter; /* Missing values to filter. */
80 bool descriptives; /* Descriptive statistics should be calculated */
81 bool quartiles; /* Quartiles should be calculated */
83 bool exact; /* Whether exact calculations have been requested */
84 double timer; /* Maximum time (in minutes) to wait for exact calculations */
87 static void one_sample_insert_variables (const struct npar_test *test,
88 struct const_hsh_table *variables);
90 static void two_sample_insert_variables (const struct npar_test *test,
91 struct const_hsh_table *variables);
96 npar_execute(struct casereader *input,
97 const struct npar_specs *specs,
98 const struct dataset *ds)
101 struct descriptives *summary_descriptives = NULL;
103 for ( t = 0 ; t < specs->n_tests; ++t )
105 const struct npar_test *test = specs->test[t];
106 if ( NULL == test->execute )
108 msg (SW, _("NPAR subcommand not currently implemented."));
111 test->execute (ds, casereader_clone (input), specs->filter, test, specs->exact, specs->timer);
114 if ( specs->descriptives )
116 summary_descriptives = xnmalloc (sizeof (*summary_descriptives),
119 npar_summary_calc_descriptives (summary_descriptives,
120 casereader_clone (input),
122 specs->vv, specs->n_vars,
126 if ( (specs->descriptives || specs->quartiles)
127 && !taint_has_tainted_successor (casereader_get_taint (input)) )
128 do_summary_box (summary_descriptives, specs->vv, specs->n_vars );
130 free (summary_descriptives);
131 casereader_destroy (input);
135 cmd_npar_tests (struct lexer *lexer, struct dataset *ds)
139 struct npar_specs npar_specs = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
140 struct const_hsh_table *var_hash;
141 struct casegrouper *grouper;
142 struct casereader *input, *group;
144 npar_specs.pool = pool_create ();
146 var_hash = const_hsh_create_pool (npar_specs.pool, 0,
147 compare_vars_by_name, hash_var_by_name,
150 if ( ! parse_npar_tests (lexer, ds, &cmd, &npar_specs) )
152 pool_destroy (npar_specs.pool);
156 for (i = 0; i < npar_specs.n_tests; ++i )
158 const struct npar_test *test = npar_specs.test[i];
159 test->insert_variables (test, var_hash);
162 npar_specs.vv = (const struct variable **) const_hsh_data (var_hash);
163 npar_specs.n_vars = const_hsh_count (var_hash);
165 if ( cmd.sbc_statistics )
169 for ( i = 0 ; i < NPAR_ST_count; ++i )
171 if ( cmd.a_statistics[i] )
175 case NPAR_ST_DESCRIPTIVES:
176 npar_specs.descriptives = true;
178 case NPAR_ST_QUARTILES:
179 npar_specs.quartiles = true;
182 npar_specs.quartiles = true;
183 npar_specs.descriptives = true;
192 npar_specs.filter = cmd.incl == NPAR_EXCLUDE ? MV_ANY : MV_SYSTEM;
194 input = proc_open (ds);
195 if ( cmd.miss == NPAR_LISTWISE )
197 input = casereader_create_filter_missing (input,
205 grouper = casegrouper_create_splits (input, dataset_dict (ds));
206 while (casegrouper_get_next_group (grouper, &group))
207 npar_execute (group, &npar_specs, ds);
208 ok = casegrouper_destroy (grouper);
209 ok = proc_commit (ds) && ok;
211 const_hsh_destroy (var_hash);
213 pool_destroy (npar_specs.pool);
215 return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
219 npar_custom_chisquare (struct lexer *lexer, struct dataset *ds,
220 struct cmd_npar_tests *cmd UNUSED, void *aux )
222 struct npar_specs *specs = aux;
224 struct chisquare_test *cstp = pool_alloc(specs->pool, sizeof(*cstp));
225 struct one_sample_test *tp = (struct one_sample_test *) cstp;
227 ((struct npar_test *)tp)->execute = chisquare_execute;
228 ((struct npar_test *)tp)->insert_variables = one_sample_insert_variables;
230 if (!parse_variables_const_pool (lexer, specs->pool, dataset_dict (ds),
231 &tp->vars, &tp->n_vars,
232 PV_NO_SCRATCH | PV_NO_DUPLICATE))
237 cstp->ranged = false;
239 if ( lex_match (lexer, '('))
242 if ( ! lex_force_num (lexer)) return 0;
243 cstp->lo = lex_integer (lexer);
245 lex_force_match (lexer, ',');
246 if (! lex_force_num (lexer) ) return 0;
247 cstp->hi = lex_integer (lexer);
248 if ( cstp->lo >= cstp->hi )
251 _("The specified value of HI (%d) is "
252 "lower than the specified value of LO (%d)"),
257 if (! lex_force_match (lexer, ')')) return 0;
260 cstp->n_expected = 0;
261 cstp->expected = NULL;
262 if ( lex_match (lexer, '/') )
264 if ( lex_match_id (lexer, "EXPECTED") )
266 lex_force_match (lexer, '=');
267 if ( ! lex_match_id (lexer, "EQUAL") )
271 while ( lex_is_number(lexer) )
275 f = lex_number (lexer);
277 if ( lex_match (lexer, '*'))
280 f = lex_number (lexer);
283 lex_match (lexer, ',');
285 cstp->n_expected += n;
286 cstp->expected = pool_realloc (specs->pool,
290 for ( i = cstp->n_expected - n ;
291 i < cstp->n_expected;
293 cstp->expected[i] = f;
299 lex_put_back (lexer, '/');
302 if ( cstp->ranged && cstp->n_expected > 0 &&
303 cstp->n_expected != cstp->hi - cstp->lo + 1 )
306 _("%d expected values were given, but the specified "
307 "range (%d-%d) requires exactly %d values."),
308 cstp->n_expected, cstp->lo, cstp->hi,
309 cstp->hi - cstp->lo +1);
314 specs->test = pool_realloc (specs->pool,
316 sizeof(*specs->test) * specs->n_tests);
318 specs->test[specs->n_tests - 1] = (struct npar_test *) tp;
325 npar_custom_binomial (struct lexer *lexer, struct dataset *ds,
326 struct cmd_npar_tests *cmd UNUSED, void *aux)
328 struct npar_specs *specs = aux;
329 struct binomial_test *btp = pool_alloc(specs->pool, sizeof(*btp));
330 struct one_sample_test *tp = (struct one_sample_test *) btp;
332 ((struct npar_test *)tp)->execute = binomial_execute;
333 ((struct npar_test *)tp)->insert_variables = one_sample_insert_variables;
335 btp->category1 = btp->category2 = btp->cutpoint = SYSMIS;
337 if ( lex_match(lexer, '(') )
339 if ( lex_force_num (lexer) )
341 btp->p = lex_number (lexer);
343 lex_force_match (lexer, ')');
349 if ( lex_match (lexer, '=') )
351 if (parse_variables_const_pool (lexer, specs->pool, dataset_dict (ds),
352 &tp->vars, &tp->n_vars,
353 PV_NUMERIC | PV_NO_SCRATCH | PV_NO_DUPLICATE) )
355 if ( lex_match (lexer, '('))
357 lex_force_num (lexer);
358 btp->category1 = lex_number (lexer);
360 if ( lex_match (lexer, ','))
362 if ( ! lex_force_num (lexer) ) return 2;
363 btp->category2 = lex_number (lexer);
368 btp->cutpoint = btp->category1;
371 lex_force_match (lexer, ')');
379 specs->test = pool_realloc (specs->pool,
381 sizeof(*specs->test) * specs->n_tests);
383 specs->test[specs->n_tests - 1] = (struct npar_test *) tp;
389 bool parse_two_sample_related_test (struct lexer *lexer,
390 const struct dictionary *dict,
391 struct cmd_npar_tests *cmd,
392 struct two_sample_test *test_parameters,
398 parse_two_sample_related_test (struct lexer *lexer,
399 const struct dictionary *dict,
400 struct cmd_npar_tests *cmd UNUSED,
401 struct two_sample_test *test_parameters,
408 const struct variable **vlist1;
411 const struct variable **vlist2;
414 ((struct npar_test *)test_parameters)->insert_variables = two_sample_insert_variables;
416 if (!parse_variables_const_pool (lexer, pool,
419 PV_NUMERIC | PV_NO_SCRATCH | PV_NO_DUPLICATE) )
422 if ( lex_match(lexer, T_WITH))
425 if ( !parse_variables_const_pool (lexer, pool, dict,
427 PV_NUMERIC | PV_NO_SCRATCH | PV_NO_DUPLICATE) )
430 paired = (lex_match (lexer, '(') &&
431 lex_match_id (lexer, "PAIRED") && lex_match (lexer, ')'));
439 if ( n_vlist1 != n_vlist2)
440 msg (SE, _("PAIRED was specified but the number of variables "
441 "preceding WITH (%zu) did not match the number "
442 "following (%zu)."), n_vlist1, n_vlist2);
444 test_parameters->n_pairs = n_vlist1 ;
448 test_parameters->n_pairs = n_vlist1 * n_vlist2;
453 test_parameters->n_pairs = (n_vlist1 * (n_vlist1 - 1)) / 2 ;
456 test_parameters->pairs =
457 pool_alloc (pool, sizeof ( variable_pair) * test_parameters->n_pairs);
464 assert (n_vlist1 == n_vlist2);
465 for ( i = 0 ; i < n_vlist1; ++i )
467 test_parameters->pairs[n][1] = vlist1[i];
468 test_parameters->pairs[n][0] = vlist2[i];
475 for ( i = 0 ; i < n_vlist1; ++i )
477 for ( j = 0 ; j < n_vlist2; ++j )
479 test_parameters->pairs[n][1] = vlist1[i];
480 test_parameters->pairs[n][0] = vlist2[j];
489 for ( i = 0 ; i < n_vlist1 - 1; ++i )
491 for ( j = i + 1 ; j < n_vlist1; ++j )
493 assert ( n < test_parameters->n_pairs);
494 test_parameters->pairs[n][1] = vlist1[i];
495 test_parameters->pairs[n][0] = vlist1[j];
501 assert ( n == test_parameters->n_pairs);
507 npar_custom_wilcoxon (struct lexer *lexer,
509 struct cmd_npar_tests *cmd, void *aux )
511 struct npar_specs *specs = aux;
513 struct two_sample_test *tp = pool_alloc (specs->pool, sizeof(*tp));
514 ((struct npar_test *)tp)->execute = wilcoxon_execute;
516 if (!parse_two_sample_related_test (lexer, dataset_dict (ds), cmd,
521 specs->test = pool_realloc (specs->pool,
523 sizeof(*specs->test) * specs->n_tests);
524 specs->test[specs->n_tests - 1] = (struct npar_test *) tp;
530 npar_custom_mcnemar (struct lexer *lexer,
532 struct cmd_npar_tests *cmd, void *aux )
534 struct npar_specs *specs = aux;
536 struct two_sample_test *tp = pool_alloc(specs->pool, sizeof(*tp));
537 ((struct npar_test *)tp)->execute = NULL;
540 if (!parse_two_sample_related_test (lexer, dataset_dict (ds),
541 cmd, tp, specs->pool) )
545 specs->test = pool_realloc (specs->pool,
547 sizeof(*specs->test) * specs->n_tests);
548 specs->test[specs->n_tests - 1] = (struct npar_test *) tp;
554 npar_custom_sign (struct lexer *lexer, struct dataset *ds,
555 struct cmd_npar_tests *cmd, void *aux )
557 struct npar_specs *specs = aux;
559 struct two_sample_test *tp = pool_alloc(specs->pool, sizeof(*tp));
560 ((struct npar_test *)tp)->execute = NULL;
563 if (!parse_two_sample_related_test (lexer, dataset_dict (ds), cmd,
568 specs->test = pool_realloc (specs->pool,
570 sizeof(*specs->test) * specs->n_tests);
571 specs->test[specs->n_tests - 1] = (struct npar_test *) tp;
576 /* Insert the variables for TEST into VAR_HASH */
578 one_sample_insert_variables (const struct npar_test *test,
579 struct const_hsh_table *var_hash)
582 struct one_sample_test *ost = (struct one_sample_test *) test;
584 for ( i = 0 ; i < ost->n_vars ; ++i )
585 const_hsh_insert (var_hash, ost->vars[i]);
589 two_sample_insert_variables (const struct npar_test *test,
590 struct const_hsh_table *var_hash)
594 const struct two_sample_test *tst = (const struct two_sample_test *) test;
596 for ( i = 0 ; i < tst->n_pairs ; ++i )
598 variable_pair *pair = &tst->pairs[i];
600 const_hsh_insert (var_hash, (*pair)[0]);
601 const_hsh_insert (var_hash, (*pair)[1]);
608 npar_custom_method (struct lexer *lexer, struct dataset *ds UNUSED,
609 struct cmd_npar_tests *test UNUSED, void *aux)
611 struct npar_specs *specs = aux;
613 if ( lex_match_id (lexer, "EXACT") )
617 if (lex_match_id (lexer, "TIMER"))
621 if ( lex_match (lexer, '('))
623 if ( lex_force_num (lexer) )
625 specs->timer = lex_number (lexer);
628 lex_force_match (lexer, ')');