1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 2004, 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 <gsl/gsl_cdf.h>
20 #include <libpspp/message.h>
25 #include <math/sort.h>
26 #include <math/order-stats.h>
27 #include <math/percentiles.h>
28 #include <math/tukey-hinges.h>
29 #include <math/box-whisker.h>
30 #include <math/trimmed-mean.h>
31 #include <math/extrema.h>
33 #include <data/case.h>
34 #include <data/casegrouper.h>
35 #include <data/casereader.h>
36 #include <data/casewriter.h>
37 #include <data/dictionary.h>
38 #include <data/procedure.h>
39 #include <data/subcase.h>
40 #include <data/value-labels.h>
41 #include <data/variable.h>
42 #include <language/command.h>
43 #include <language/dictionary/split-file.h>
44 #include <language/lexer/lexer.h>
45 #include <libpspp/compiler.h>
46 #include <libpspp/hash.h>
47 #include <libpspp/message.h>
48 #include <libpspp/misc.h>
49 #include <libpspp/str.h>
50 #include <math/moments.h>
51 #include <output/chart-provider.h>
52 #include <output/charts/box-whisker.h>
53 #include <output/charts/cartesian.h>
54 #include <output/manager.h>
55 #include <output/table.h>
61 #define _(msgid) gettext (msgid)
62 #define N_(msgid) msgid
65 #include <output/chart.h>
66 #include <output/charts/plot-hist.h>
67 #include <output/charts/plot-chart.h>
68 #include <math/histogram.h>
75 missing=miss:pairwise/!listwise,
77 incl:include/!exclude;
78 +compare=cmp:variables/!groups;
81 +plot[plt_]=stemleaf,boxplot,npplot,:spreadlevel(*d:n),histogram,all,none;
83 +statistics[st_]=descriptives,:extreme(*d:n),all,none.
91 static struct cmd_examine cmd;
93 static const struct variable **dependent_vars;
94 static size_t n_dependent_vars;
98 static subc_list_double percentile_list;
99 static enum pc_alg percentile_algorithm;
101 struct factor_metrics
103 struct moments1 *moments;
105 struct percentile **ptl;
108 struct statistic *tukey_hinges;
109 struct statistic *box_whisker;
110 struct statistic *trimmed_mean;
111 struct statistic *histogram;
112 struct order_stats *np;
114 /* Three quartiles indexing into PTL */
115 struct percentile **quartiles;
117 /* A reader sorted in ASCENDING order */
118 struct casereader *up_reader;
120 /* The minimum value of all the weights */
123 /* Sum of all weights, including those for missing values */
126 /* Sum of weights of non_missing values */
139 struct extrema *minima;
140 struct extrema *maxima;
147 union value value[2];
149 /* An array of factor metrics, one for each variable */
150 struct factor_metrics *metrics;
155 /* We need to make a list of this structure */
158 /* The independent variable */
159 const struct variable const* indep_var[2];
161 /* A list of results for this factor */
162 struct ll_list result_list ;
167 factor_destroy (struct xfactor *fctr)
169 struct ll *ll = ll_head (&fctr->result_list);
170 while (ll != ll_null (&fctr->result_list))
173 struct factor_result *result =
174 ll_data (ll, struct factor_result, ll);
177 for (v = 0; v < n_dependent_vars; ++v)
180 moments1_destroy (result->metrics[v].moments);
181 extrema_destroy (result->metrics[v].minima);
182 extrema_destroy (result->metrics[v].maxima);
183 statistic_destroy (result->metrics[v].trimmed_mean);
184 statistic_destroy (result->metrics[v].tukey_hinges);
185 statistic_destroy (result->metrics[v].box_whisker);
186 statistic_destroy (result->metrics[v].histogram);
187 for (i = 0 ; i < result->metrics[v].n_ptiles; ++i)
188 statistic_destroy ((struct statistic *) result->metrics[v].ptl[i]);
189 free (result->metrics[v].ptl);
190 free (result->metrics[v].quartiles);
191 casereader_destroy (result->metrics[v].up_reader);
194 for (i = 0; i < 2; i++)
195 if (fctr->indep_var[i])
196 value_destroy (&result->value[i],
197 var_get_width (fctr->indep_var[i]));
198 free (result->metrics);
204 static struct xfactor level0_factor;
205 static struct ll_list factor_list;
207 /* Parse the clause specifying the factors */
208 static int examine_parse_independent_vars (struct lexer *lexer,
209 const struct dictionary *dict,
210 struct cmd_examine *cmd);
212 /* Output functions */
213 static void show_summary (const struct variable **dependent_var, int n_dep_var,
214 const struct dictionary *dict,
215 const struct xfactor *f);
218 static void show_descriptives (const struct variable **dependent_var,
220 const struct xfactor *f);
223 static void show_percentiles (const struct variable **dependent_var,
225 const struct xfactor *f);
228 static void show_extremes (const struct variable **dependent_var,
230 const struct xfactor *f);
235 /* Per Split function */
236 static void run_examine (struct cmd_examine *, struct casereader *,
239 static void output_examine (const struct dictionary *dict);
242 void factor_calc (const struct ccase *c, int case_no,
243 double weight, bool case_missing);
246 /* Represent a factor as a string, so it can be
247 printed in a human readable fashion */
248 static void factor_to_string (const struct xfactor *fctr,
249 const struct factor_result *result,
252 /* Represent a factor as a string, so it can be
253 printed in a human readable fashion,
254 but sacrificing some readablility for the sake of brevity */
256 factor_to_string_concise (const struct xfactor *fctr,
257 const struct factor_result *result,
263 /* Categories of missing values to exclude. */
264 static enum mv_class exclude_values;
267 cmd_examine (struct lexer *lexer, struct dataset *ds)
269 struct casegrouper *grouper;
270 struct casereader *group;
273 subc_list_double_create (&percentile_list);
274 percentile_algorithm = PC_HAVERAGE;
276 ll_init (&factor_list);
278 if ( !parse_examine (lexer, ds, &cmd, NULL) )
280 subc_list_double_destroy (&percentile_list);
284 /* If /MISSING=INCLUDE is set, then user missing values are ignored */
285 exclude_values = cmd.incl == XMN_INCLUDE ? MV_SYSTEM : MV_ANY;
287 if ( cmd.st_n == SYSMIS )
290 if ( ! cmd.sbc_cinterval)
291 cmd.n_cinterval[0] = 95.0;
293 /* If descriptives have been requested, make sure the
294 quartiles are calculated */
295 if ( cmd.a_statistics[XMN_ST_DESCRIPTIVES] )
297 subc_list_double_push (&percentile_list, 25);
298 subc_list_double_push (&percentile_list, 50);
299 subc_list_double_push (&percentile_list, 75);
302 grouper = casegrouper_create_splits (proc_open (ds), dataset_dict (ds));
304 while (casegrouper_get_next_group (grouper, &group))
306 struct casereader *reader =
307 casereader_create_arithmetic_sequence (group, 1, 1);
309 run_examine (&cmd, reader, ds);
312 ok = casegrouper_destroy (grouper);
313 ok = proc_commit (ds) && ok;
315 if ( dependent_vars )
316 free (dependent_vars);
318 subc_list_double_destroy (&percentile_list);
320 return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
328 struct casereader *data;
330 /* Copied directly from struct np. */
332 double dns_min, dns_max;
335 double slope, intercept;
336 double y_first, y_last;
337 double x_lower, x_upper;
341 static const struct chart_class np_plot_chart_class;
342 static const struct chart_class dnp_plot_chart_class;
344 /* Plot the normal and detrended normal plots for RESULT.
345 Label the plots with LABEL */
347 np_plot (struct np *np, const char *label)
349 struct np_plot_chart *np_plot, *dnp_plot;
353 msg (MW, _("Not creating plot because data set is empty."));
357 np_plot = xmalloc (sizeof *np_plot);
358 chart_init (&np_plot->chart, &np_plot_chart_class);
359 np_plot->label = xstrdup (label);
360 np_plot->data = casewriter_make_reader (np->writer);
361 np_plot->y_min = np->y_min;
362 np_plot->y_max = np->y_max;
363 np_plot->dns_min = np->dns_min;
364 np_plot->dns_max = np->dns_max;
366 /* Slope and intercept of the ideal normal probability line. */
367 np_plot->slope = 1.0 / np->stddev;
368 np_plot->intercept = -np->mean / np->stddev;
370 np_plot->y_first = gsl_cdf_ugaussian_Pinv (1 / (np->n + 1));
371 np_plot->y_last = gsl_cdf_ugaussian_Pinv (np->n / (np->n + 1));
373 /* Need to make sure that both the scatter plot and the ideal fit into the
375 np_plot->x_lower = MIN (
376 np->y_min, (np_plot->y_first - np_plot->intercept) / np_plot->slope);
377 np_plot->x_upper = MAX (
378 np->y_max, (np_plot->y_last - np_plot->intercept) / np_plot->slope) ;
379 np_plot->slack = (np_plot->x_upper - np_plot->x_lower) * 0.05 ;
381 dnp_plot = xmemdup (np_plot, sizeof *np_plot);
382 chart_init (&dnp_plot->chart, &dnp_plot_chart_class);
383 dnp_plot->label = xstrdup (dnp_plot->label);
384 dnp_plot->data = casereader_clone (dnp_plot->data);
386 chart_submit (&np_plot->chart);
387 chart_submit (&dnp_plot->chart);
391 np_plot_chart_draw (const struct chart *chart, plPlotter *lp)
393 const struct np_plot_chart *plot = (struct np_plot_chart *) chart;
394 struct chart_geometry geom;
395 struct casereader *data;
398 chart_geometry_init (lp, &geom);
399 chart_write_title (lp, &geom, _("Normal Q-Q Plot of %s"), plot->label);
400 chart_write_xlabel (lp, &geom, _("Observed Value"));
401 chart_write_ylabel (lp, &geom, _("Expected Normal"));
402 chart_write_xscale (lp, &geom,
403 plot->x_lower - plot->slack,
404 plot->x_upper + plot->slack, 5);
405 chart_write_yscale (lp, &geom, plot->y_first, plot->y_last, 5);
407 data = casereader_clone (plot->data);
408 for (; (c = casereader_read (data)) != NULL; case_unref (c))
409 chart_datum (lp, &geom, 0,
410 case_data_idx (c, NP_IDX_Y)->f,
411 case_data_idx (c, NP_IDX_NS)->f);
412 casereader_destroy (data);
414 chart_line (lp, &geom, plot->slope, plot->intercept,
415 plot->y_first, plot->y_last, CHART_DIM_Y);
417 chart_geometry_free (lp);
421 dnp_plot_chart_draw (const struct chart *chart, plPlotter *lp)
423 const struct np_plot_chart *plot = (struct np_plot_chart *) chart;
424 struct chart_geometry geom;
425 struct casereader *data;
428 chart_geometry_init (lp, &geom);
429 chart_write_title (lp, &geom, _("Detrended Normal Q-Q Plot of %s"),
431 chart_write_xlabel (lp, &geom, _("Observed Value"));
432 chart_write_ylabel (lp, &geom, _("Dev from Normal"));
433 chart_write_xscale (lp, &geom, plot->y_min, plot->y_max, 5);
434 chart_write_yscale (lp, &geom, plot->dns_min, plot->dns_max, 5);
436 data = casereader_clone (plot->data);
437 for (; (c = casereader_read (data)) != NULL; case_unref (c))
438 chart_datum (lp, &geom, 0, case_data_idx (c, NP_IDX_Y)->f,
439 case_data_idx (c, NP_IDX_DNS)->f);
440 casereader_destroy (data);
442 chart_line (lp, &geom, 0, 0, plot->y_min, plot->y_max, CHART_DIM_X);
444 chart_geometry_free (lp);
448 np_plot_chart_destroy (struct chart *chart)
450 struct np_plot_chart *plot = (struct np_plot_chart *) chart;
452 casereader_destroy (plot->data);
457 static const struct chart_class np_plot_chart_class =
460 np_plot_chart_destroy
463 static const struct chart_class dnp_plot_chart_class =
466 np_plot_chart_destroy
471 show_npplot (const struct variable **dependent_var,
473 const struct xfactor *fctr)
477 for (v = 0; v < n_dep_var; ++v)
480 for (ll = ll_head (&fctr->result_list);
481 ll != ll_null (&fctr->result_list);
485 const struct factor_result *result =
486 ll_data (ll, struct factor_result, ll);
488 ds_init_empty (&str);
489 ds_put_format (&str, "%s ", var_get_name (dependent_var[v]));
491 factor_to_string (fctr, result, &str);
493 np_plot ((struct np*) result->metrics[v].np, ds_cstr(&str));
495 statistic_destroy ((struct statistic *)result->metrics[v].np);
504 show_histogram (const struct variable **dependent_var,
506 const struct xfactor *fctr)
510 for (v = 0; v < n_dep_var; ++v)
513 for (ll = ll_head (&fctr->result_list);
514 ll != ll_null (&fctr->result_list);
518 const struct factor_result *result =
519 ll_data (ll, struct factor_result, ll);
520 struct histogram *histogram;
523 histogram = (struct histogram *) result->metrics[v].histogram;
524 if (histogram == NULL)
526 /* Probably all values are SYSMIS. */
530 ds_init_empty (&str);
531 ds_put_format (&str, "%s ", var_get_name (dependent_var[v]));
533 factor_to_string (fctr, result, &str);
535 moments1_calculate ((struct moments1 *) result->metrics[v].moments,
536 &n, &mean, &var, NULL, NULL);
537 chart_submit (histogram_chart_create (histogram, ds_cstr (&str),
538 n, mean, sqrt (var), false));
548 show_boxplot_groups (const struct variable **dependent_var,
550 const struct xfactor *fctr)
554 for (v = 0; v < n_dep_var; ++v)
556 const struct factor_result *result;
557 struct boxplot *boxplot;
558 double y_min = DBL_MAX;
559 double y_max = -DBL_MAX;
562 ll_for_each (result, struct factor_result, ll, &fctr->result_list)
564 struct factor_metrics *metrics = &result->metrics[v];
565 const struct ll_list *max_list = extrema_list (metrics->maxima);
566 const struct ll_list *min_list = extrema_list (metrics->minima);
567 const struct extremum *max, *min;
569 if ( ll_is_empty (max_list))
571 msg (MW, _("Not creating plot because data set is empty."));
575 max = ll_data (ll_head(max_list), struct extremum, ll);
576 min = ll_data (ll_head (min_list), struct extremum, ll);
578 y_max = MAX (y_max, max->value);
579 y_min = MIN (y_min, min->value);
582 if (fctr->indep_var[0])
583 title = xasprintf (_("Boxplot of %s vs. %s"),
584 var_to_string (dependent_var[v]),
585 var_to_string (fctr->indep_var[0]));
587 title = xasprintf (_("Boxplot of %s"),
588 var_to_string (dependent_var[v]));
589 boxplot = boxplot_create (y_min, y_max, title);
592 ll_for_each (result, struct factor_result, ll, &fctr->result_list)
594 struct factor_metrics *metrics = &result->metrics[v];
595 struct string str = DS_EMPTY_INITIALIZER;
596 factor_to_string_concise (fctr, result, &str);
597 boxplot_add_box (boxplot,
598 (struct box_whisker *) metrics->box_whisker,
600 metrics->box_whisker = NULL;
604 chart_submit (boxplot_get_chart (boxplot));
611 show_boxplot_variables (const struct variable **dependent_var,
613 const struct xfactor *fctr
617 const struct factor_result *result;
620 ll_for_each (result, struct factor_result, ll, &fctr->result_list)
623 double y_min = DBL_MAX;
624 double y_max = -DBL_MAX;
625 struct boxplot *boxplot;
627 for (v = 0; v < n_dep_var; ++v)
629 const struct factor_metrics *metrics = &result->metrics[v];
630 const struct ll *max_ll = ll_head (extrema_list (metrics->maxima));
631 const struct ll *min_ll = ll_head (extrema_list (metrics->minima));
632 const struct extremum *max = ll_data (max_ll, struct extremum, ll);
633 const struct extremum *min = ll_data (min_ll, struct extremum, ll);
635 y_max = MAX (y_max, max->value);
636 y_min = MIN (y_min, min->value);
639 ds_init_empty (&title);
640 factor_to_string (fctr, result, &title);
641 boxplot = boxplot_create (y_min, y_max, ds_cstr (&title));
644 for (v = 0; v < n_dep_var; ++v)
646 struct factor_metrics *metrics = &result->metrics[v];
647 boxplot_add_box (boxplot,
648 (struct box_whisker *) metrics->box_whisker,
649 var_get_name (dependent_var[v]));
650 metrics->box_whisker = NULL;
653 chart_submit (boxplot_get_chart (boxplot));
658 /* Show all the appropriate tables */
660 output_examine (const struct dictionary *dict)
664 show_summary (dependent_vars, n_dependent_vars, dict, &level0_factor);
666 if ( cmd.a_statistics[XMN_ST_EXTREME] )
667 show_extremes (dependent_vars, n_dependent_vars, &level0_factor);
669 if ( cmd.a_statistics[XMN_ST_DESCRIPTIVES] )
670 show_descriptives (dependent_vars, n_dependent_vars, &level0_factor);
672 if ( cmd.sbc_percentiles)
673 show_percentiles (dependent_vars, n_dependent_vars, &level0_factor);
677 if (cmd.a_plot[XMN_PLT_BOXPLOT])
678 show_boxplot_groups (dependent_vars, n_dependent_vars, &level0_factor);
680 if (cmd.a_plot[XMN_PLT_HISTOGRAM])
681 show_histogram (dependent_vars, n_dependent_vars, &level0_factor);
683 if (cmd.a_plot[XMN_PLT_NPPLOT])
684 show_npplot (dependent_vars, n_dependent_vars, &level0_factor);
687 for (ll = ll_head (&factor_list);
688 ll != ll_null (&factor_list); ll = ll_next (ll))
690 struct xfactor *factor = ll_data (ll, struct xfactor, ll);
691 show_summary (dependent_vars, n_dependent_vars, dict, factor);
693 if ( cmd.a_statistics[XMN_ST_EXTREME] )
694 show_extremes (dependent_vars, n_dependent_vars, factor);
696 if ( cmd.a_statistics[XMN_ST_DESCRIPTIVES] )
697 show_descriptives (dependent_vars, n_dependent_vars, factor);
699 if ( cmd.sbc_percentiles)
700 show_percentiles (dependent_vars, n_dependent_vars, factor);
702 if (cmd.a_plot[XMN_PLT_BOXPLOT])
704 if (cmd.cmp == XMN_GROUPS)
705 show_boxplot_groups (dependent_vars, n_dependent_vars, factor);
706 else if (cmd.cmp == XMN_VARIABLES)
707 show_boxplot_variables (dependent_vars, n_dependent_vars, factor);
710 if (cmd.a_plot[XMN_PLT_HISTOGRAM])
711 show_histogram (dependent_vars, n_dependent_vars, factor);
713 if (cmd.a_plot[XMN_PLT_NPPLOT])
714 show_npplot (dependent_vars, n_dependent_vars, factor);
718 /* Parse the PERCENTILES subcommand */
720 xmn_custom_percentiles (struct lexer *lexer, struct dataset *ds UNUSED,
721 struct cmd_examine *p UNUSED, void *aux UNUSED)
723 lex_match (lexer, '=');
725 lex_match (lexer, '(');
727 while ( lex_is_number (lexer) )
729 subc_list_double_push (&percentile_list, lex_number (lexer));
733 lex_match (lexer, ',') ;
735 lex_match (lexer, ')');
737 lex_match (lexer, '=');
739 if ( lex_match_id (lexer, "HAVERAGE"))
740 percentile_algorithm = PC_HAVERAGE;
742 else if ( lex_match_id (lexer, "WAVERAGE"))
743 percentile_algorithm = PC_WAVERAGE;
745 else if ( lex_match_id (lexer, "ROUND"))
746 percentile_algorithm = PC_ROUND;
748 else if ( lex_match_id (lexer, "EMPIRICAL"))
749 percentile_algorithm = PC_EMPIRICAL;
751 else if ( lex_match_id (lexer, "AEMPIRICAL"))
752 percentile_algorithm = PC_AEMPIRICAL;
754 else if ( lex_match_id (lexer, "NONE"))
755 percentile_algorithm = PC_NONE;
758 if ( 0 == subc_list_double_count (&percentile_list))
760 subc_list_double_push (&percentile_list, 5);
761 subc_list_double_push (&percentile_list, 10);
762 subc_list_double_push (&percentile_list, 25);
763 subc_list_double_push (&percentile_list, 50);
764 subc_list_double_push (&percentile_list, 75);
765 subc_list_double_push (&percentile_list, 90);
766 subc_list_double_push (&percentile_list, 95);
772 /* TOTAL and NOTOTAL are simple, mutually exclusive flags */
774 xmn_custom_total (struct lexer *lexer UNUSED, struct dataset *ds UNUSED,
775 struct cmd_examine *p, void *aux UNUSED)
777 if ( p->sbc_nototal )
779 msg (SE, _("%s and %s are mutually exclusive"),"TOTAL","NOTOTAL");
787 xmn_custom_nototal (struct lexer *lexer UNUSED, struct dataset *ds UNUSED,
788 struct cmd_examine *p, void *aux UNUSED)
792 msg (SE, _("%s and %s are mutually exclusive"), "TOTAL", "NOTOTAL");
801 /* Parser for the variables sub command
802 Returns 1 on success */
804 xmn_custom_variables (struct lexer *lexer, struct dataset *ds,
805 struct cmd_examine *cmd,
808 const struct dictionary *dict = dataset_dict (ds);
809 lex_match (lexer, '=');
811 if ( (lex_token (lexer) != T_ID || dict_lookup_var (dict, lex_tokid (lexer)) == NULL)
812 && lex_token (lexer) != T_ALL)
817 if (!parse_variables_const (lexer, dict, &dependent_vars, &n_dependent_vars,
818 PV_NO_DUPLICATE | PV_NUMERIC | PV_NO_SCRATCH) )
820 free (dependent_vars);
824 assert (n_dependent_vars);
827 if ( lex_match (lexer, T_BY))
830 success = examine_parse_independent_vars (lexer, dict, cmd);
833 free (dependent_vars);
843 /* Parse the clause specifying the factors */
845 examine_parse_independent_vars (struct lexer *lexer,
846 const struct dictionary *dict,
847 struct cmd_examine *cmd)
850 struct xfactor *sf = xmalloc (sizeof *sf);
852 ll_init (&sf->result_list);
854 if ( (lex_token (lexer) != T_ID ||
855 dict_lookup_var (dict, lex_tokid (lexer)) == NULL)
856 && lex_token (lexer) != T_ALL)
862 sf->indep_var[0] = parse_variable (lexer, dict);
863 sf->indep_var[1] = NULL;
865 if ( lex_token (lexer) == T_BY )
867 lex_match (lexer, T_BY);
869 if ( (lex_token (lexer) != T_ID ||
870 dict_lookup_var (dict, lex_tokid (lexer)) == NULL)
871 && lex_token (lexer) != T_ALL)
877 sf->indep_var[1] = parse_variable (lexer, dict);
879 ll_push_tail (&factor_list, &sf->ll);
882 ll_push_tail (&factor_list, &sf->ll);
884 lex_match (lexer, ',');
886 if ( lex_token (lexer) == '.' || lex_token (lexer) == '/' )
889 success = examine_parse_independent_vars (lexer, dict, cmd);
898 examine_group (struct cmd_examine *cmd, struct casereader *reader, int level,
899 const struct dictionary *dict, struct xfactor *factor)
902 const struct variable *wv = dict_get_weight (dict);
905 struct factor_result *result = xzalloc (sizeof (*result));
908 for (i = 0; i < 2; i++)
909 if (factor->indep_var[i])
910 value_init (&result->value[i], var_get_width (factor->indep_var[i]));
912 result->metrics = xcalloc (n_dependent_vars, sizeof (*result->metrics));
914 if ( cmd->a_statistics[XMN_ST_EXTREME] )
915 n_extrema = cmd->st_n;
918 c = casereader_peek (reader, 0);
922 for (i = 0; i < 2; i++)
923 if (factor->indep_var[i])
924 value_copy (&result->value[i], case_data (c, factor->indep_var[i]),
925 var_get_width (factor->indep_var[i]));
929 for (v = 0; v < n_dependent_vars; ++v)
931 struct casewriter *writer;
932 struct casereader *input = casereader_clone (reader);
934 result->metrics[v].moments = moments1_create (MOMENT_KURTOSIS);
935 result->metrics[v].minima = extrema_create (n_extrema, EXTREME_MINIMA);
936 result->metrics[v].maxima = extrema_create (n_extrema, EXTREME_MAXIMA);
937 result->metrics[v].cmin = DBL_MAX;
939 if (cmd->a_statistics[XMN_ST_DESCRIPTIVES] ||
940 cmd->a_plot[XMN_PLT_BOXPLOT] ||
941 cmd->a_plot[XMN_PLT_NPPLOT] ||
942 cmd->sbc_percentiles)
944 /* In this case, we need to sort the data, so we create a sorting
946 struct subcase up_ordering;
947 subcase_init_var (&up_ordering, dependent_vars[v], SC_ASCEND);
948 writer = sort_create_writer (&up_ordering,
949 casereader_get_proto (reader));
950 subcase_destroy (&up_ordering);
954 /* but in this case, sorting is unnecessary, so an ordinary
955 casewriter is sufficient */
957 autopaging_writer_create (casereader_get_proto (reader));
961 /* Sort or just iterate, whilst calculating moments etc */
962 while ((c = casereader_read (input)) != NULL)
964 int n_vals = caseproto_get_n_widths (casereader_get_proto (reader));
965 const casenumber loc = case_data_idx (c, n_vals - 1)->f;
967 const double weight = wv ? case_data (c, wv)->f : 1.0;
968 const union value *value = case_data (c, dependent_vars[v]);
970 if (weight != SYSMIS)
971 minimize (&result->metrics[v].cmin, weight);
973 moments1_add (result->metrics[v].moments,
977 result->metrics[v].n += weight;
979 if ( ! var_is_value_missing (dependent_vars[v], value, MV_ANY) )
980 result->metrics[v].n_valid += weight;
982 extrema_add (result->metrics[v].maxima,
987 extrema_add (result->metrics[v].minima,
992 casewriter_write (writer, c);
994 casereader_destroy (input);
995 result->metrics[v].up_reader = casewriter_make_reader (writer);
998 /* If percentiles or descriptives have been requested, then a
999 second pass through the data (which has now been sorted)
1001 if ( cmd->a_statistics[XMN_ST_DESCRIPTIVES] ||
1002 cmd->a_plot[XMN_PLT_BOXPLOT] ||
1003 cmd->a_plot[XMN_PLT_NPPLOT] ||
1004 cmd->sbc_percentiles)
1006 for (v = 0; v < n_dependent_vars; ++v)
1010 struct order_stats **os ;
1011 struct factor_metrics *metric = &result->metrics[v];
1013 metric->n_ptiles = percentile_list.n_data;
1015 metric->ptl = xcalloc (metric->n_ptiles,
1016 sizeof (struct percentile *));
1018 metric->quartiles = xcalloc (3, sizeof (*metric->quartiles));
1020 for (i = 0 ; i < metric->n_ptiles; ++i)
1022 metric->ptl[i] = (struct percentile *)
1023 percentile_create (percentile_list.data[i] / 100.0, metric->n_valid);
1025 if ( percentile_list.data[i] == 25)
1026 metric->quartiles[0] = metric->ptl[i];
1027 else if ( percentile_list.data[i] == 50)
1028 metric->quartiles[1] = metric->ptl[i];
1029 else if ( percentile_list.data[i] == 75)
1030 metric->quartiles[2] = metric->ptl[i];
1033 metric->tukey_hinges = tukey_hinges_create (metric->n_valid, metric->cmin);
1034 metric->trimmed_mean = trimmed_mean_create (metric->n_valid, 0.05);
1036 n_os = metric->n_ptiles + 2;
1038 if ( cmd->a_plot[XMN_PLT_NPPLOT] )
1040 metric->np = np_create (metric->moments);
1044 os = xcalloc (sizeof (struct order_stats *), n_os);
1046 for (i = 0 ; i < metric->n_ptiles ; ++i )
1048 os[i] = (struct order_stats *) metric->ptl[i];
1051 os[i] = (struct order_stats *) metric->tukey_hinges;
1052 os[i+1] = (struct order_stats *) metric->trimmed_mean;
1054 if (cmd->a_plot[XMN_PLT_NPPLOT])
1055 os[i+2] = metric->np;
1057 order_stats_accumulate (os, n_os,
1058 casereader_clone (metric->up_reader),
1059 wv, dependent_vars[v], MV_ANY);
1064 /* FIXME: Do this in the above loop */
1065 if ( cmd->a_plot[XMN_PLT_HISTOGRAM] )
1068 struct casereader *input = casereader_clone (reader);
1070 for (v = 0; v < n_dependent_vars; ++v)
1072 const struct extremum *max, *min;
1073 struct factor_metrics *metric = &result->metrics[v];
1075 const struct ll_list *max_list =
1076 extrema_list (result->metrics[v].maxima);
1078 const struct ll_list *min_list =
1079 extrema_list (result->metrics[v].minima);
1081 if ( ll_is_empty (max_list))
1083 msg (MW, _("Not creating plot because data set is empty."));
1087 assert (! ll_is_empty (min_list));
1089 max = (const struct extremum *)
1090 ll_data (ll_head(max_list), struct extremum, ll);
1092 min = (const struct extremum *)
1093 ll_data (ll_head (min_list), struct extremum, ll);
1095 metric->histogram = histogram_create (10, min->value, max->value);
1098 while ((c = casereader_read (input)) != NULL)
1100 const double weight = wv ? case_data (c, wv)->f : 1.0;
1102 for (v = 0; v < n_dependent_vars; ++v)
1104 struct factor_metrics *metric = &result->metrics[v];
1105 if ( metric->histogram)
1106 histogram_add ((struct histogram *) metric->histogram,
1107 case_data (c, dependent_vars[v])->f, weight);
1111 casereader_destroy (input);
1114 /* In this case, a third iteration is required */
1115 if (cmd->a_plot[XMN_PLT_BOXPLOT])
1117 for (v = 0; v < n_dependent_vars; ++v)
1119 struct factor_metrics *metric = &result->metrics[v];
1120 int n_vals = caseproto_get_n_widths (casereader_get_proto (
1121 metric->up_reader));
1123 metric->box_whisker =
1124 box_whisker_create ((struct tukey_hinges *) metric->tukey_hinges,
1125 cmd->v_id, n_vals - 1);
1127 order_stats_accumulate ((struct order_stats **) &metric->box_whisker,
1129 casereader_clone (metric->up_reader),
1130 wv, dependent_vars[v], MV_ANY);
1134 ll_push_tail (&factor->result_list, &result->ll);
1135 casereader_destroy (reader);
1140 run_examine (struct cmd_examine *cmd, struct casereader *input,
1144 const struct dictionary *dict = dataset_dict (ds);
1146 struct casereader *level0 = casereader_clone (input);
1148 c = casereader_peek (input, 0);
1151 casereader_destroy (input);
1155 output_split_file_values (ds, c);
1158 ll_init (&level0_factor.result_list);
1160 examine_group (cmd, level0, 0, dict, &level0_factor);
1162 for (ll = ll_head (&factor_list);
1163 ll != ll_null (&factor_list);
1166 struct xfactor *factor = ll_data (ll, struct xfactor, ll);
1168 struct casereader *group = NULL;
1169 struct casereader *level1;
1170 struct casegrouper *grouper1 = NULL;
1172 level1 = casereader_clone (input);
1173 level1 = sort_execute_1var (level1, factor->indep_var[0]);
1174 grouper1 = casegrouper_create_vars (level1, &factor->indep_var[0], 1);
1176 while (casegrouper_get_next_group (grouper1, &group))
1178 struct casereader *group_copy = casereader_clone (group);
1180 if ( !factor->indep_var[1])
1181 examine_group (cmd, group_copy, 1, dict, factor);
1185 struct casereader *group2 = NULL;
1186 struct casegrouper *grouper2 = NULL;
1188 group_copy = sort_execute_1var (group_copy,
1189 factor->indep_var[1]);
1191 grouper2 = casegrouper_create_vars (group_copy,
1192 &factor->indep_var[1], 1);
1194 while (casegrouper_get_next_group (grouper2, &group2))
1196 examine_group (cmd, group2, 2, dict, factor);
1199 casegrouper_destroy (grouper2);
1202 casereader_destroy (group);
1204 casegrouper_destroy (grouper1);
1207 casereader_destroy (input);
1209 output_examine (dict);
1211 factor_destroy (&level0_factor);
1215 for (ll = ll_head (&factor_list);
1216 ll != ll_null (&factor_list);
1219 struct xfactor *f = ll_data (ll, struct xfactor, ll);
1228 show_summary (const struct variable **dependent_var, int n_dep_var,
1229 const struct dictionary *dict,
1230 const struct xfactor *fctr)
1232 const struct variable *wv = dict_get_weight (dict);
1233 const struct fmt_spec *wfmt = wv ? var_get_print_format (wv) : & F_8_0;
1235 static const char *subtitle[]=
1243 int heading_columns = 1;
1245 const int heading_rows = 3;
1246 struct tab_table *tbl;
1253 if ( fctr->indep_var[0] )
1255 heading_columns = 2;
1257 if ( fctr->indep_var[1] )
1259 heading_columns = 3;
1263 n_rows *= ll_count (&fctr->result_list);
1264 n_rows += heading_rows;
1266 n_cols = heading_columns + 6;
1268 tbl = tab_create (n_cols, n_rows, 0);
1269 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
1271 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
1273 /* Outline the box */
1278 n_cols - 1, n_rows - 1);
1280 /* Vertical lines for the data only */
1285 n_cols - 1, n_rows - 1);
1288 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
1289 tab_hline (tbl, TAL_1, heading_columns, n_cols - 1, 1 );
1290 tab_hline (tbl, TAL_1, heading_columns, n_cols - 1, heading_rows -1 );
1292 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
1295 tab_title (tbl, _("Case Processing Summary"));
1297 tab_joint_text (tbl, heading_columns, 0,
1299 TAB_CENTER | TAT_TITLE,
1302 /* Remove lines ... */
1309 for (j = 0 ; j < 3 ; ++j)
1311 tab_text (tbl, heading_columns + j * 2 , 2, TAB_CENTER | TAT_TITLE,
1314 tab_text (tbl, heading_columns + j * 2 + 1, 2, TAB_CENTER | TAT_TITLE,
1317 tab_joint_text (tbl, heading_columns + j * 2 , 1,
1318 heading_columns + j * 2 + 1, 1,
1319 TAB_CENTER | TAT_TITLE,
1322 tab_box (tbl, -1, -1,
1324 heading_columns + j * 2, 1,
1325 heading_columns + j * 2 + 1, 1);
1329 /* Titles for the independent variables */
1330 if ( fctr->indep_var[0] )
1332 tab_text (tbl, 1, heading_rows - 1, TAB_CENTER | TAT_TITLE,
1333 var_to_string (fctr->indep_var[0]));
1335 if ( fctr->indep_var[1] )
1337 tab_text (tbl, 2, heading_rows - 1, TAB_CENTER | TAT_TITLE,
1338 var_to_string (fctr->indep_var[1]));
1342 for (v = 0 ; v < n_dep_var ; ++v)
1346 const union value *last_value = NULL;
1349 tab_hline (tbl, TAL_1, 0, n_cols -1 ,
1350 v * ll_count (&fctr->result_list)
1355 v * ll_count (&fctr->result_list) + heading_rows,
1356 TAB_LEFT | TAT_TITLE,
1357 var_to_string (dependent_var[v])
1361 for (ll = ll_head (&fctr->result_list);
1362 ll != ll_null (&fctr->result_list); ll = ll_next (ll))
1365 const struct factor_result *result =
1366 ll_data (ll, struct factor_result, ll);
1368 if ( fctr->indep_var[0] )
1371 if ( last_value == NULL ||
1372 !value_equal (last_value, &result->value[0],
1373 var_get_width (fctr->indep_var[0])))
1377 last_value = &result->value[0];
1378 ds_init_empty (&str);
1380 var_append_value_name (fctr->indep_var[0], &result->value[0],
1385 v * ll_count (&fctr->result_list),
1386 TAB_LEFT | TAT_TITLE,
1391 if ( fctr->indep_var[1] && j > 0)
1392 tab_hline (tbl, TAL_1, 1, n_cols - 1,
1394 v * ll_count (&fctr->result_list));
1397 if ( fctr->indep_var[1])
1401 ds_init_empty (&str);
1403 var_append_value_name (fctr->indep_var[1],
1404 &result->value[1], &str);
1408 v * ll_count (&fctr->result_list),
1409 TAB_LEFT | TAT_TITLE,
1417 moments1_calculate (result->metrics[v].moments,
1418 &n, &result->metrics[v].mean,
1419 &result->metrics[v].variance,
1420 &result->metrics[v].skewness,
1421 &result->metrics[v].kurtosis);
1423 result->metrics[v].se_mean = sqrt (result->metrics[v].variance / n) ;
1426 tab_double (tbl, heading_columns,
1427 heading_rows + j + v * ll_count (&fctr->result_list),
1431 tab_text (tbl, heading_columns + 1,
1432 heading_rows + j + v * ll_count (&fctr->result_list),
1433 TAB_RIGHT | TAT_PRINTF,
1434 "%g%%", n * 100.0 / result->metrics[v].n);
1437 tab_double (tbl, heading_columns + 2,
1438 heading_rows + j + v * ll_count (&fctr->result_list),
1440 result->metrics[v].n - n,
1443 tab_text (tbl, heading_columns + 3,
1444 heading_rows + j + v * ll_count (&fctr->result_list),
1445 TAB_RIGHT | TAT_PRINTF,
1447 (result->metrics[v].n - n) * 100.0 / result->metrics[v].n
1450 /* Total Valid + Missing */
1451 tab_double (tbl, heading_columns + 4,
1452 heading_rows + j + v * ll_count (&fctr->result_list),
1454 result->metrics[v].n,
1457 tab_text (tbl, heading_columns + 5,
1458 heading_rows + j + v * ll_count (&fctr->result_list),
1459 TAB_RIGHT | TAT_PRINTF,
1461 (result->metrics[v].n) * 100.0 / result->metrics[v].n
1472 #define DESCRIPTIVE_ROWS 13
1475 show_descriptives (const struct variable **dependent_var,
1477 const struct xfactor *fctr)
1480 int heading_columns = 3;
1482 const int heading_rows = 1;
1483 struct tab_table *tbl;
1490 if ( fctr->indep_var[0] )
1492 heading_columns = 4;
1494 if ( fctr->indep_var[1] )
1496 heading_columns = 5;
1500 n_rows *= ll_count (&fctr->result_list) * DESCRIPTIVE_ROWS;
1501 n_rows += heading_rows;
1503 n_cols = heading_columns + 2;
1505 tbl = tab_create (n_cols, n_rows, 0);
1506 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
1508 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
1510 /* Outline the box */
1515 n_cols - 1, n_rows - 1);
1518 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
1519 tab_hline (tbl, TAL_2, 1, n_cols - 1, heading_rows );
1521 tab_vline (tbl, TAL_1, n_cols - 1, 0, n_rows - 1);
1524 if ( fctr->indep_var[0])
1525 tab_text (tbl, 1, 0, TAT_TITLE, var_to_string (fctr->indep_var[0]));
1527 if ( fctr->indep_var[1])
1528 tab_text (tbl, 2, 0, TAT_TITLE, var_to_string (fctr->indep_var[1]));
1530 for (v = 0 ; v < n_dep_var ; ++v )
1535 const int row_var_start =
1536 v * DESCRIPTIVE_ROWS * ll_count(&fctr->result_list);
1540 heading_rows + row_var_start,
1541 TAB_LEFT | TAT_TITLE,
1542 var_to_string (dependent_var[v])
1545 for (ll = ll_head (&fctr->result_list);
1546 ll != ll_null (&fctr->result_list); i++, ll = ll_next (ll))
1548 const struct factor_result *result =
1549 ll_data (ll, struct factor_result, ll);
1552 gsl_cdf_tdist_Qinv ((1 - cmd.n_cinterval[0] / 100.0) / 2.0,
1553 result->metrics[v].n - 1);
1555 if ( i > 0 || v > 0 )
1557 const int left_col = (i == 0) ? 0 : 1;
1558 tab_hline (tbl, TAL_1, left_col, n_cols - 1,
1559 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS);
1562 if ( fctr->indep_var[0])
1565 ds_init_empty (&vstr);
1566 var_append_value_name (fctr->indep_var[0],
1567 &result->value[0], &vstr);
1570 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1579 tab_text (tbl, n_cols - 4,
1580 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1584 tab_text (tbl, n_cols - 4,
1585 heading_rows + row_var_start + 1 + i * DESCRIPTIVE_ROWS,
1586 TAB_LEFT | TAT_PRINTF,
1587 _("%g%% Confidence Interval for Mean"),
1588 cmd.n_cinterval[0]);
1590 tab_text (tbl, n_cols - 3,
1591 heading_rows + row_var_start + 1 + i * DESCRIPTIVE_ROWS,
1595 tab_text (tbl, n_cols - 3,
1596 heading_rows + row_var_start + 2 + i * DESCRIPTIVE_ROWS,
1600 tab_text (tbl, n_cols - 4,
1601 heading_rows + row_var_start + 3 + i * DESCRIPTIVE_ROWS,
1602 TAB_LEFT | TAT_PRINTF,
1603 _("5%% Trimmed Mean"));
1605 tab_text (tbl, n_cols - 4,
1606 heading_rows + row_var_start + 4 + i * DESCRIPTIVE_ROWS,
1610 tab_text (tbl, n_cols - 4,
1611 heading_rows + row_var_start + 5 + i * DESCRIPTIVE_ROWS,
1615 tab_text (tbl, n_cols - 4,
1616 heading_rows + row_var_start + 6 + i * DESCRIPTIVE_ROWS,
1618 _("Std. Deviation"));
1620 tab_text (tbl, n_cols - 4,
1621 heading_rows + row_var_start + 7 + i * DESCRIPTIVE_ROWS,
1625 tab_text (tbl, n_cols - 4,
1626 heading_rows + row_var_start + 8 + i * DESCRIPTIVE_ROWS,
1630 tab_text (tbl, n_cols - 4,
1631 heading_rows + row_var_start + 9 + i * DESCRIPTIVE_ROWS,
1635 tab_text (tbl, n_cols - 4,
1636 heading_rows + row_var_start + 10 + i * DESCRIPTIVE_ROWS,
1638 _("Interquartile Range"));
1641 tab_text (tbl, n_cols - 4,
1642 heading_rows + row_var_start + 11 + i * DESCRIPTIVE_ROWS,
1646 tab_text (tbl, n_cols - 4,
1647 heading_rows + row_var_start + 12 + i * DESCRIPTIVE_ROWS,
1652 /* Now the statistics ... */
1654 tab_double (tbl, n_cols - 2,
1655 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1657 result->metrics[v].mean,
1660 tab_double (tbl, n_cols - 1,
1661 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1663 result->metrics[v].se_mean,
1667 tab_double (tbl, n_cols - 2,
1668 heading_rows + row_var_start + 1 + i * DESCRIPTIVE_ROWS,
1670 result->metrics[v].mean - t *
1671 result->metrics[v].se_mean,
1674 tab_double (tbl, n_cols - 2,
1675 heading_rows + row_var_start + 2 + i * DESCRIPTIVE_ROWS,
1677 result->metrics[v].mean + t *
1678 result->metrics[v].se_mean,
1682 tab_double (tbl, n_cols - 2,
1683 heading_rows + row_var_start + 3 + i * DESCRIPTIVE_ROWS,
1685 trimmed_mean_calculate ((struct trimmed_mean *) result->metrics[v].trimmed_mean),
1689 tab_double (tbl, n_cols - 2,
1690 heading_rows + row_var_start + 4 + i * DESCRIPTIVE_ROWS,
1692 percentile_calculate (result->metrics[v].quartiles[1], percentile_algorithm),
1696 tab_double (tbl, n_cols - 2,
1697 heading_rows + row_var_start + 5 + i * DESCRIPTIVE_ROWS,
1699 result->metrics[v].variance,
1702 tab_double (tbl, n_cols - 2,
1703 heading_rows + row_var_start + 6 + i * DESCRIPTIVE_ROWS,
1705 sqrt (result->metrics[v].variance),
1708 tab_double (tbl, n_cols - 2,
1709 heading_rows + row_var_start + 10 + i * DESCRIPTIVE_ROWS,
1711 percentile_calculate (result->metrics[v].quartiles[2],
1712 percentile_algorithm) -
1713 percentile_calculate (result->metrics[v].quartiles[0],
1714 percentile_algorithm),
1718 tab_double (tbl, n_cols - 2,
1719 heading_rows + row_var_start + 11 + i * DESCRIPTIVE_ROWS,
1721 result->metrics[v].skewness,
1724 tab_double (tbl, n_cols - 2,
1725 heading_rows + row_var_start + 12 + i * DESCRIPTIVE_ROWS,
1727 result->metrics[v].kurtosis,
1730 tab_double (tbl, n_cols - 1,
1731 heading_rows + row_var_start + 11 + i * DESCRIPTIVE_ROWS,
1733 calc_seskew (result->metrics[v].n),
1736 tab_double (tbl, n_cols - 1,
1737 heading_rows + row_var_start + 12 + i * DESCRIPTIVE_ROWS,
1739 calc_sekurt (result->metrics[v].n),
1743 struct extremum *minimum, *maximum ;
1745 struct ll *max_ll = ll_head (extrema_list (result->metrics[v].maxima));
1746 struct ll *min_ll = ll_head (extrema_list (result->metrics[v].minima));
1748 maximum = ll_data (max_ll, struct extremum, ll);
1749 minimum = ll_data (min_ll, struct extremum, ll);
1751 tab_double (tbl, n_cols - 2,
1752 heading_rows + row_var_start + 7 + i * DESCRIPTIVE_ROWS,
1757 tab_double (tbl, n_cols - 2,
1758 heading_rows + row_var_start + 8 + i * DESCRIPTIVE_ROWS,
1763 tab_double (tbl, n_cols - 2,
1764 heading_rows + row_var_start + 9 + i * DESCRIPTIVE_ROWS,
1766 maximum->value - minimum->value,
1772 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
1774 tab_title (tbl, _("Descriptives"));
1776 tab_text (tbl, n_cols - 2, 0, TAB_CENTER | TAT_TITLE,
1779 tab_text (tbl, n_cols - 1, 0, TAB_CENTER | TAT_TITLE,
1788 show_extremes (const struct variable **dependent_var,
1790 const struct xfactor *fctr)
1793 int heading_columns = 3;
1795 const int heading_rows = 1;
1796 struct tab_table *tbl;
1803 if ( fctr->indep_var[0] )
1805 heading_columns = 4;
1807 if ( fctr->indep_var[1] )
1809 heading_columns = 5;
1813 n_rows *= ll_count (&fctr->result_list) * cmd.st_n * 2;
1814 n_rows += heading_rows;
1816 n_cols = heading_columns + 2;
1818 tbl = tab_create (n_cols, n_rows, 0);
1819 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
1821 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
1823 /* Outline the box */
1828 n_cols - 1, n_rows - 1);
1831 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
1832 tab_hline (tbl, TAL_2, 1, n_cols - 1, heading_rows );
1833 tab_vline (tbl, TAL_1, n_cols - 1, 0, n_rows - 1);
1835 if ( fctr->indep_var[0])
1836 tab_text (tbl, 1, 0, TAT_TITLE, var_to_string (fctr->indep_var[0]));
1838 if ( fctr->indep_var[1])
1839 tab_text (tbl, 2, 0, TAT_TITLE, var_to_string (fctr->indep_var[1]));
1841 for (v = 0 ; v < n_dep_var ; ++v )
1845 const int row_var_start = v * cmd.st_n * 2 * ll_count(&fctr->result_list);
1849 heading_rows + row_var_start,
1850 TAB_LEFT | TAT_TITLE,
1851 var_to_string (dependent_var[v])
1854 for (ll = ll_head (&fctr->result_list);
1855 ll != ll_null (&fctr->result_list); i++, ll = ll_next (ll))
1860 const int row_result_start = i * cmd.st_n * 2;
1862 const struct factor_result *result =
1863 ll_data (ll, struct factor_result, ll);
1866 tab_hline (tbl, TAL_1, 1, n_cols - 1,
1867 heading_rows + row_var_start + row_result_start);
1869 tab_hline (tbl, TAL_1, heading_columns - 2, n_cols - 1,
1870 heading_rows + row_var_start + row_result_start + cmd.st_n);
1872 for ( e = 1; e <= cmd.st_n; ++e )
1874 tab_text (tbl, n_cols - 3,
1875 heading_rows + row_var_start + row_result_start + e - 1,
1876 TAB_RIGHT | TAT_PRINTF,
1879 tab_text (tbl, n_cols - 3,
1880 heading_rows + row_var_start + row_result_start + cmd.st_n + e - 1,
1881 TAB_RIGHT | TAT_PRINTF,
1886 min_ll = ll_head (extrema_list (result->metrics[v].minima));
1887 for (e = 0; e < cmd.st_n;)
1889 struct extremum *minimum = ll_data (min_ll, struct extremum, ll);
1890 double weight = minimum->weight;
1892 while (weight-- > 0 && e < cmd.st_n)
1894 tab_double (tbl, n_cols - 1,
1895 heading_rows + row_var_start + row_result_start + cmd.st_n + e,
1901 tab_fixed (tbl, n_cols - 2,
1902 heading_rows + row_var_start +
1903 row_result_start + cmd.st_n + e,
1910 min_ll = ll_next (min_ll);
1914 max_ll = ll_head (extrema_list (result->metrics[v].maxima));
1915 for (e = 0; e < cmd.st_n;)
1917 struct extremum *maximum = ll_data (max_ll, struct extremum, ll);
1918 double weight = maximum->weight;
1920 while (weight-- > 0 && e < cmd.st_n)
1922 tab_double (tbl, n_cols - 1,
1923 heading_rows + row_var_start +
1924 row_result_start + e,
1930 tab_fixed (tbl, n_cols - 2,
1931 heading_rows + row_var_start +
1932 row_result_start + e,
1939 max_ll = ll_next (max_ll);
1943 if ( fctr->indep_var[0])
1946 ds_init_empty (&vstr);
1947 var_append_value_name (fctr->indep_var[0],
1948 &result->value[0], &vstr);
1951 heading_rows + row_var_start + row_result_start,
1960 tab_text (tbl, n_cols - 4,
1961 heading_rows + row_var_start + row_result_start,
1965 tab_text (tbl, n_cols - 4,
1966 heading_rows + row_var_start + row_result_start + cmd.st_n,
1972 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
1975 tab_title (tbl, _("Extreme Values"));
1978 tab_text (tbl, n_cols - 2, 0, TAB_CENTER | TAT_TITLE,
1982 tab_text (tbl, n_cols - 1, 0, TAB_CENTER | TAT_TITLE,
1988 #define PERCENTILE_ROWS 2
1991 show_percentiles (const struct variable **dependent_var,
1993 const struct xfactor *fctr)
1997 int heading_columns = 2;
1999 const int n_percentiles = subc_list_double_count (&percentile_list);
2000 const int heading_rows = 2;
2001 struct tab_table *tbl;
2008 if ( fctr->indep_var[0] )
2010 heading_columns = 3;
2012 if ( fctr->indep_var[1] )
2014 heading_columns = 4;
2018 n_rows *= ll_count (&fctr->result_list) * PERCENTILE_ROWS;
2019 n_rows += heading_rows;
2021 n_cols = heading_columns + n_percentiles;
2023 tbl = tab_create (n_cols, n_rows, 0);
2024 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
2026 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
2028 /* Outline the box */
2033 n_cols - 1, n_rows - 1);
2036 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
2037 tab_hline (tbl, TAL_2, 1, n_cols - 1, heading_rows );
2039 if ( fctr->indep_var[0])
2040 tab_text (tbl, 1, 1, TAT_TITLE, var_to_string (fctr->indep_var[0]));
2042 if ( fctr->indep_var[1])
2043 tab_text (tbl, 2, 1, TAT_TITLE, var_to_string (fctr->indep_var[1]));
2045 for (v = 0 ; v < n_dep_var ; ++v )
2051 const int row_var_start =
2052 v * PERCENTILE_ROWS * ll_count(&fctr->result_list);
2056 heading_rows + row_var_start,
2057 TAB_LEFT | TAT_TITLE,
2058 var_to_string (dependent_var[v])
2061 for (ll = ll_head (&fctr->result_list);
2062 ll != ll_null (&fctr->result_list); i++, ll = ll_next (ll))
2065 const struct factor_result *result =
2066 ll_data (ll, struct factor_result, ll);
2068 if ( i > 0 || v > 0 )
2070 const int left_col = (i == 0) ? 0 : 1;
2071 tab_hline (tbl, TAL_1, left_col, n_cols - 1,
2072 heading_rows + row_var_start + i * PERCENTILE_ROWS);
2075 if ( fctr->indep_var[0])
2078 ds_init_empty (&vstr);
2079 var_append_value_name (fctr->indep_var[0],
2080 &result->value[0], &vstr);
2083 heading_rows + row_var_start + i * PERCENTILE_ROWS,
2092 tab_text (tbl, n_cols - n_percentiles - 1,
2093 heading_rows + row_var_start + i * PERCENTILE_ROWS,
2095 ptile_alg_desc [percentile_algorithm]);
2098 tab_text (tbl, n_cols - n_percentiles - 1,
2099 heading_rows + row_var_start + 1 + i * PERCENTILE_ROWS,
2101 _("Tukey's Hinges"));
2104 tab_vline (tbl, TAL_1, n_cols - n_percentiles -1, heading_rows, n_rows - 1);
2106 tukey_hinges_calculate ((struct tukey_hinges *) result->metrics[v].tukey_hinges,
2109 for (j = 0; j < n_percentiles; ++j)
2111 double hinge = SYSMIS;
2112 tab_double (tbl, n_cols - n_percentiles + j,
2113 heading_rows + row_var_start + i * PERCENTILE_ROWS,
2115 percentile_calculate (result->metrics[v].ptl[j],
2116 percentile_algorithm),
2120 if ( result->metrics[v].ptl[j]->ptile == 0.5)
2122 else if ( result->metrics[v].ptl[j]->ptile == 0.25)
2124 else if ( result->metrics[v].ptl[j]->ptile == 0.75)
2127 if ( hinge != SYSMIS)
2128 tab_double (tbl, n_cols - n_percentiles + j,
2129 heading_rows + row_var_start + 1 + i * PERCENTILE_ROWS,
2139 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
2141 tab_title (tbl, _("Percentiles"));
2144 for (i = 0 ; i < n_percentiles; ++i )
2146 tab_text (tbl, n_cols - n_percentiles + i, 1,
2147 TAB_CENTER | TAT_TITLE | TAT_PRINTF,
2149 subc_list_double_at (&percentile_list, i)
2155 tab_joint_text (tbl,
2156 n_cols - n_percentiles, 0,
2158 TAB_CENTER | TAT_TITLE,
2161 /* Vertical lines for the data only */
2165 n_cols - n_percentiles, 1,
2166 n_cols - 1, n_rows - 1);
2168 tab_hline (tbl, TAL_1, n_cols - n_percentiles, n_cols - 1, 1);
2176 factor_to_string_concise (const struct xfactor *fctr,
2177 const struct factor_result *result,
2181 if (fctr->indep_var[0])
2183 var_append_value_name (fctr->indep_var[0], &result->value[0], str);
2185 if ( fctr->indep_var[1] )
2187 ds_put_cstr (str, ",");
2189 var_append_value_name (fctr->indep_var[1], &result->value[1], str);
2191 ds_put_cstr (str, ")");
2198 factor_to_string (const struct xfactor *fctr,
2199 const struct factor_result *result,
2203 if (fctr->indep_var[0])
2205 ds_put_format (str, "(%s = ", var_get_name (fctr->indep_var[0]));
2207 var_append_value_name (fctr->indep_var[0], &result->value[0], str);
2209 if ( fctr->indep_var[1] )
2211 ds_put_cstr (str, ",");
2212 ds_put_format (str, "%s = ", var_get_name (fctr->indep_var[1]));
2214 var_append_value_name (fctr->indep_var[1], &result->value[1], str);
2216 ds_put_cstr (str, ")");