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/charts/box-whisker.h>
52 #include <output/charts/cartesian.h>
53 #include <output/manager.h>
54 #include <output/table.h>
60 #define _(msgid) gettext (msgid)
61 #define N_(msgid) msgid
64 #include <output/chart.h>
65 #include <output/charts/plot-hist.h>
66 #include <output/charts/plot-chart.h>
67 #include <math/histogram.h>
74 missing=miss:pairwise/!listwise,
76 incl:include/!exclude;
77 +compare=cmp:variables/!groups;
80 +plot[plt_]=stemleaf,boxplot,npplot,:spreadlevel(*d:n),histogram,all,none;
82 +statistics[st_]=descriptives,:extreme(*d:n),all,none.
90 static struct cmd_examine cmd;
92 static const struct variable **dependent_vars;
93 static size_t n_dependent_vars;
97 static subc_list_double percentile_list;
98 static enum pc_alg percentile_algorithm;
100 struct factor_metrics
102 struct moments1 *moments;
104 struct percentile **ptl;
107 struct statistic *tukey_hinges;
108 struct statistic *box_whisker;
109 struct statistic *trimmed_mean;
110 struct statistic *histogram;
111 struct order_stats *np;
113 /* Three quartiles indexing into PTL */
114 struct percentile **quartiles;
116 /* A reader sorted in ASCENDING order */
117 struct casereader *up_reader;
119 /* The minimum value of all the weights */
122 /* Sum of all weights, including those for missing values */
125 /* Sum of weights of non_missing values */
138 struct extrema *minima;
139 struct extrema *maxima;
146 union value value[2];
148 /* An array of factor metrics, one for each variable */
149 struct factor_metrics *metrics;
154 /* We need to make a list of this structure */
157 /* The independent variable */
158 const struct variable const* indep_var[2];
160 /* A list of results for this factor */
161 struct ll_list result_list ;
166 factor_destroy (struct xfactor *fctr)
168 struct ll *ll = ll_head (&fctr->result_list);
169 while (ll != ll_null (&fctr->result_list))
172 struct factor_result *result =
173 ll_data (ll, struct factor_result, ll);
176 for (v = 0; v < n_dependent_vars; ++v)
179 moments1_destroy (result->metrics[v].moments);
180 extrema_destroy (result->metrics[v].minima);
181 extrema_destroy (result->metrics[v].maxima);
182 statistic_destroy (result->metrics[v].trimmed_mean);
183 statistic_destroy (result->metrics[v].tukey_hinges);
184 statistic_destroy (result->metrics[v].box_whisker);
185 statistic_destroy (result->metrics[v].histogram);
186 for (i = 0 ; i < result->metrics[v].n_ptiles; ++i)
187 statistic_destroy ((struct statistic *) result->metrics[v].ptl[i]);
188 free (result->metrics[v].ptl);
189 free (result->metrics[v].quartiles);
190 casereader_destroy (result->metrics[v].up_reader);
193 for (i = 0; i < 2; i++)
194 if (fctr->indep_var[i])
195 value_destroy (&result->value[i],
196 var_get_width (fctr->indep_var[i]));
197 free (result->metrics);
203 static struct xfactor level0_factor;
204 static struct ll_list factor_list;
206 /* Parse the clause specifying the factors */
207 static int examine_parse_independent_vars (struct lexer *lexer,
208 const struct dictionary *dict,
209 struct cmd_examine *cmd);
211 /* Output functions */
212 static void show_summary (const struct variable **dependent_var, int n_dep_var,
213 const struct dictionary *dict,
214 const struct xfactor *f);
217 static void show_descriptives (const struct variable **dependent_var,
219 const struct xfactor *f);
222 static void show_percentiles (const struct variable **dependent_var,
224 const struct xfactor *f);
227 static void show_extremes (const struct variable **dependent_var,
229 const struct xfactor *f);
234 /* Per Split function */
235 static void run_examine (struct cmd_examine *, struct casereader *,
238 static void output_examine (const struct dictionary *dict);
241 void factor_calc (const struct ccase *c, int case_no,
242 double weight, bool case_missing);
245 /* Represent a factor as a string, so it can be
246 printed in a human readable fashion */
247 static void factor_to_string (const struct xfactor *fctr,
248 const struct factor_result *result,
251 /* Represent a factor as a string, so it can be
252 printed in a human readable fashion,
253 but sacrificing some readablility for the sake of brevity */
255 factor_to_string_concise (const struct xfactor *fctr,
256 const struct factor_result *result,
262 /* Categories of missing values to exclude. */
263 static enum mv_class exclude_values;
266 cmd_examine (struct lexer *lexer, struct dataset *ds)
268 struct casegrouper *grouper;
269 struct casereader *group;
272 subc_list_double_create (&percentile_list);
273 percentile_algorithm = PC_HAVERAGE;
275 ll_init (&factor_list);
277 if ( !parse_examine (lexer, ds, &cmd, NULL) )
279 subc_list_double_destroy (&percentile_list);
283 /* If /MISSING=INCLUDE is set, then user missing values are ignored */
284 exclude_values = cmd.incl == XMN_INCLUDE ? MV_SYSTEM : MV_ANY;
286 if ( cmd.st_n == SYSMIS )
289 if ( ! cmd.sbc_cinterval)
290 cmd.n_cinterval[0] = 95.0;
292 /* If descriptives have been requested, make sure the
293 quartiles are calculated */
294 if ( cmd.a_statistics[XMN_ST_DESCRIPTIVES] )
296 subc_list_double_push (&percentile_list, 25);
297 subc_list_double_push (&percentile_list, 50);
298 subc_list_double_push (&percentile_list, 75);
301 grouper = casegrouper_create_splits (proc_open (ds), dataset_dict (ds));
303 while (casegrouper_get_next_group (grouper, &group))
305 struct casereader *reader =
306 casereader_create_arithmetic_sequence (group, 1, 1);
308 run_examine (&cmd, reader, ds);
311 ok = casegrouper_destroy (grouper);
312 ok = proc_commit (ds) && ok;
314 if ( dependent_vars )
315 free (dependent_vars);
317 subc_list_double_destroy (&percentile_list);
319 return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
323 /* Plot the normal and detrended normal plots for RESULT.
324 Label the plots with LABEL */
326 np_plot (struct np *np, const char *label)
328 double yfirst = 0, ylast = 0;
335 struct chart *np_chart;
337 /* Detrended Normal Plot */
338 struct chart *dnp_chart;
340 /* The slope and intercept of the ideal normal probability line */
341 const double slope = 1.0 / np->stddev;
342 const double intercept = -np->mean / np->stddev;
346 msg (MW, _("Not creating plot because data set is empty."));
350 np_chart = chart_create ();
351 dnp_chart = chart_create ();
353 if ( !np_chart || ! dnp_chart )
356 chart_write_title (np_chart, _("Normal Q-Q Plot of %s"), label);
357 chart_write_xlabel (np_chart, _("Observed Value"));
358 chart_write_ylabel (np_chart, _("Expected Normal"));
360 chart_write_title (dnp_chart, _("Detrended Normal Q-Q Plot of %s"),
362 chart_write_xlabel (dnp_chart, _("Observed Value"));
363 chart_write_ylabel (dnp_chart, _("Dev from Normal"));
365 yfirst = gsl_cdf_ugaussian_Pinv (1 / (np->n + 1));
366 ylast = gsl_cdf_ugaussian_Pinv (np->n / (np->n + 1));
368 /* Need to make sure that both the scatter plot and the ideal fit into the
370 x_lower = MIN (np->y_min, (yfirst - intercept) / slope) ;
371 x_upper = MAX (np->y_max, (ylast - intercept) / slope) ;
372 slack = (x_upper - x_lower) * 0.05 ;
374 chart_write_xscale (np_chart, x_lower - slack, x_upper + slack, 5);
375 chart_write_xscale (dnp_chart, np->y_min, np->y_max, 5);
377 chart_write_yscale (np_chart, yfirst, ylast, 5);
378 chart_write_yscale (dnp_chart, np->dns_min, np->dns_max, 5);
381 struct casereader *reader = casewriter_make_reader (np->writer);
383 while ((c = casereader_read (reader)) != NULL)
385 chart_datum (np_chart, 0, case_data_idx (c, NP_IDX_Y)->f, case_data_idx (c, NP_IDX_NS)->f);
386 chart_datum (dnp_chart, 0, case_data_idx (c, NP_IDX_Y)->f, case_data_idx (c, NP_IDX_DNS)->f);
390 casereader_destroy (reader);
393 chart_line (dnp_chart, 0, 0, np->y_min, np->y_max , CHART_DIM_X);
394 chart_line (np_chart, slope, intercept, yfirst, ylast , CHART_DIM_Y);
396 chart_submit (np_chart);
397 chart_submit (dnp_chart);
402 show_npplot (const struct variable **dependent_var,
404 const struct xfactor *fctr)
408 for (v = 0; v < n_dep_var; ++v)
411 for (ll = ll_head (&fctr->result_list);
412 ll != ll_null (&fctr->result_list);
416 const struct factor_result *result =
417 ll_data (ll, struct factor_result, ll);
419 ds_init_empty (&str);
420 ds_put_format (&str, "%s ", var_get_name (dependent_var[v]));
422 factor_to_string (fctr, result, &str);
424 np_plot ((struct np*) result->metrics[v].np, ds_cstr(&str));
426 statistic_destroy ((struct statistic *)result->metrics[v].np);
435 show_histogram (const struct variable **dependent_var,
437 const struct xfactor *fctr)
441 for (v = 0; v < n_dep_var; ++v)
444 for (ll = ll_head (&fctr->result_list);
445 ll != ll_null (&fctr->result_list);
449 const struct factor_result *result =
450 ll_data (ll, struct factor_result, ll);
452 ds_init_empty (&str);
453 ds_put_format (&str, "%s ", var_get_name (dependent_var[v]));
455 factor_to_string (fctr, result, &str);
457 histogram_plot ((struct histogram *) result->metrics[v].histogram,
459 (struct moments1 *) result->metrics[v].moments);
469 show_boxplot_groups (const struct variable **dependent_var,
471 const struct xfactor *fctr)
475 for (v = 0; v < n_dep_var; ++v)
479 struct chart *ch = chart_create ();
480 double y_min = DBL_MAX;
481 double y_max = -DBL_MAX;
483 for (ll = ll_head (&fctr->result_list);
484 ll != ll_null (&fctr->result_list);
487 const struct extremum *max, *min;
488 const struct factor_result *result =
489 ll_data (ll, struct factor_result, ll);
491 const struct ll_list *max_list =
492 extrema_list (result->metrics[v].maxima);
494 const struct ll_list *min_list =
495 extrema_list (result->metrics[v].minima);
497 if ( ll_is_empty (max_list))
499 msg (MW, _("Not creating plot because data set is empty."));
503 max = (const struct extremum *)
504 ll_data (ll_head(max_list), struct extremum, ll);
506 min = (const struct extremum *)
507 ll_data (ll_head (min_list), struct extremum, ll);
509 y_max = MAX (y_max, max->value);
510 y_min = MIN (y_min, min->value);
513 boxplot_draw_yscale (ch, y_max, y_min);
515 if ( fctr->indep_var[0])
516 chart_write_title (ch, _("Boxplot of %s vs. %s"),
517 var_to_string (dependent_var[v]),
518 var_to_string (fctr->indep_var[0]) );
520 chart_write_title (ch, _("Boxplot of %s"),
521 var_to_string (dependent_var[v]));
523 for (ll = ll_head (&fctr->result_list);
524 ll != ll_null (&fctr->result_list);
527 const struct factor_result *result =
528 ll_data (ll, struct factor_result, ll);
531 const double box_width = (ch->data_right - ch->data_left)
532 / (ll_count (&fctr->result_list) * 2.0 ) ;
534 const double box_centre = (f++ * 2 + 1) * box_width + ch->data_left;
536 ds_init_empty (&str);
537 factor_to_string_concise (fctr, result, &str);
539 boxplot_draw_boxplot (ch,
540 box_centre, box_width,
541 (const struct box_whisker *)
542 result->metrics[v].box_whisker,
555 show_boxplot_variables (const struct variable **dependent_var,
557 const struct xfactor *fctr
563 const struct ll_list *result_list = &fctr->result_list;
565 for (ll = ll_head (result_list);
566 ll != ll_null (result_list);
571 struct chart *ch = chart_create ();
572 double y_min = DBL_MAX;
573 double y_max = -DBL_MAX;
575 const struct factor_result *result =
576 ll_data (ll, struct factor_result, ll);
578 const double box_width = (ch->data_right - ch->data_left)
579 / (n_dep_var * 2.0 ) ;
581 for (v = 0; v < n_dep_var; ++v)
583 const struct ll *max_ll =
584 ll_head (extrema_list (result->metrics[v].maxima));
585 const struct ll *min_ll =
586 ll_head (extrema_list (result->metrics[v].minima));
588 const struct extremum *max =
589 (const struct extremum *) ll_data (max_ll, struct extremum, ll);
591 const struct extremum *min =
592 (const struct extremum *) ll_data (min_ll, struct extremum, ll);
594 y_max = MAX (y_max, max->value);
595 y_min = MIN (y_min, min->value);
599 boxplot_draw_yscale (ch, y_max, y_min);
601 ds_init_empty (&title);
602 factor_to_string (fctr, result, &title);
605 ds_put_format (&title, "%s = ", var_get_name (fctr->indep_var[0]));
606 var_append_value_name (fctr->indep_var[0], &result->value[0], &title);
609 chart_write_title (ch, ds_cstr (&title));
612 for (v = 0; v < n_dep_var; ++v)
615 const double box_centre = (v * 2 + 1) * box_width + ch->data_left;
617 ds_init_empty (&str);
618 ds_init_cstr (&str, var_get_name (dependent_var[v]));
620 boxplot_draw_boxplot (ch,
621 box_centre, box_width,
622 (const struct box_whisker *) result->metrics[v].box_whisker,
633 /* Show all the appropriate tables */
635 output_examine (const struct dictionary *dict)
639 show_summary (dependent_vars, n_dependent_vars, dict, &level0_factor);
641 if ( cmd.a_statistics[XMN_ST_EXTREME] )
642 show_extremes (dependent_vars, n_dependent_vars, &level0_factor);
644 if ( cmd.a_statistics[XMN_ST_DESCRIPTIVES] )
645 show_descriptives (dependent_vars, n_dependent_vars, &level0_factor);
647 if ( cmd.sbc_percentiles)
648 show_percentiles (dependent_vars, n_dependent_vars, &level0_factor);
652 if (cmd.a_plot[XMN_PLT_BOXPLOT])
653 show_boxplot_groups (dependent_vars, n_dependent_vars, &level0_factor);
655 if (cmd.a_plot[XMN_PLT_HISTOGRAM])
656 show_histogram (dependent_vars, n_dependent_vars, &level0_factor);
658 if (cmd.a_plot[XMN_PLT_NPPLOT])
659 show_npplot (dependent_vars, n_dependent_vars, &level0_factor);
662 for (ll = ll_head (&factor_list);
663 ll != ll_null (&factor_list); ll = ll_next (ll))
665 struct xfactor *factor = ll_data (ll, struct xfactor, ll);
666 show_summary (dependent_vars, n_dependent_vars, dict, factor);
668 if ( cmd.a_statistics[XMN_ST_EXTREME] )
669 show_extremes (dependent_vars, n_dependent_vars, factor);
671 if ( cmd.a_statistics[XMN_ST_DESCRIPTIVES] )
672 show_descriptives (dependent_vars, n_dependent_vars, factor);
674 if ( cmd.sbc_percentiles)
675 show_percentiles (dependent_vars, n_dependent_vars, factor);
677 if (cmd.a_plot[XMN_PLT_BOXPLOT] &&
678 cmd.cmp == XMN_GROUPS)
679 show_boxplot_groups (dependent_vars, n_dependent_vars, factor);
682 if (cmd.a_plot[XMN_PLT_BOXPLOT] &&
683 cmd.cmp == XMN_VARIABLES)
684 show_boxplot_variables (dependent_vars, n_dependent_vars,
687 if (cmd.a_plot[XMN_PLT_HISTOGRAM])
688 show_histogram (dependent_vars, n_dependent_vars, factor);
690 if (cmd.a_plot[XMN_PLT_NPPLOT])
691 show_npplot (dependent_vars, n_dependent_vars, factor);
695 /* Parse the PERCENTILES subcommand */
697 xmn_custom_percentiles (struct lexer *lexer, struct dataset *ds UNUSED,
698 struct cmd_examine *p UNUSED, void *aux UNUSED)
700 lex_match (lexer, '=');
702 lex_match (lexer, '(');
704 while ( lex_is_number (lexer) )
706 subc_list_double_push (&percentile_list, lex_number (lexer));
710 lex_match (lexer, ',') ;
712 lex_match (lexer, ')');
714 lex_match (lexer, '=');
716 if ( lex_match_id (lexer, "HAVERAGE"))
717 percentile_algorithm = PC_HAVERAGE;
719 else if ( lex_match_id (lexer, "WAVERAGE"))
720 percentile_algorithm = PC_WAVERAGE;
722 else if ( lex_match_id (lexer, "ROUND"))
723 percentile_algorithm = PC_ROUND;
725 else if ( lex_match_id (lexer, "EMPIRICAL"))
726 percentile_algorithm = PC_EMPIRICAL;
728 else if ( lex_match_id (lexer, "AEMPIRICAL"))
729 percentile_algorithm = PC_AEMPIRICAL;
731 else if ( lex_match_id (lexer, "NONE"))
732 percentile_algorithm = PC_NONE;
735 if ( 0 == subc_list_double_count (&percentile_list))
737 subc_list_double_push (&percentile_list, 5);
738 subc_list_double_push (&percentile_list, 10);
739 subc_list_double_push (&percentile_list, 25);
740 subc_list_double_push (&percentile_list, 50);
741 subc_list_double_push (&percentile_list, 75);
742 subc_list_double_push (&percentile_list, 90);
743 subc_list_double_push (&percentile_list, 95);
749 /* TOTAL and NOTOTAL are simple, mutually exclusive flags */
751 xmn_custom_total (struct lexer *lexer UNUSED, struct dataset *ds UNUSED,
752 struct cmd_examine *p, void *aux UNUSED)
754 if ( p->sbc_nototal )
756 msg (SE, _("%s and %s are mutually exclusive"),"TOTAL","NOTOTAL");
764 xmn_custom_nototal (struct lexer *lexer UNUSED, struct dataset *ds UNUSED,
765 struct cmd_examine *p, void *aux UNUSED)
769 msg (SE, _("%s and %s are mutually exclusive"), "TOTAL", "NOTOTAL");
778 /* Parser for the variables sub command
779 Returns 1 on success */
781 xmn_custom_variables (struct lexer *lexer, struct dataset *ds,
782 struct cmd_examine *cmd,
785 const struct dictionary *dict = dataset_dict (ds);
786 lex_match (lexer, '=');
788 if ( (lex_token (lexer) != T_ID || dict_lookup_var (dict, lex_tokid (lexer)) == NULL)
789 && lex_token (lexer) != T_ALL)
794 if (!parse_variables_const (lexer, dict, &dependent_vars, &n_dependent_vars,
795 PV_NO_DUPLICATE | PV_NUMERIC | PV_NO_SCRATCH) )
797 free (dependent_vars);
801 assert (n_dependent_vars);
804 if ( lex_match (lexer, T_BY))
807 success = examine_parse_independent_vars (lexer, dict, cmd);
810 free (dependent_vars);
820 /* Parse the clause specifying the factors */
822 examine_parse_independent_vars (struct lexer *lexer,
823 const struct dictionary *dict,
824 struct cmd_examine *cmd)
827 struct xfactor *sf = xmalloc (sizeof *sf);
829 ll_init (&sf->result_list);
831 if ( (lex_token (lexer) != T_ID ||
832 dict_lookup_var (dict, lex_tokid (lexer)) == NULL)
833 && lex_token (lexer) != T_ALL)
839 sf->indep_var[0] = parse_variable (lexer, dict);
840 sf->indep_var[1] = NULL;
842 if ( lex_token (lexer) == T_BY )
844 lex_match (lexer, T_BY);
846 if ( (lex_token (lexer) != T_ID ||
847 dict_lookup_var (dict, lex_tokid (lexer)) == NULL)
848 && lex_token (lexer) != T_ALL)
854 sf->indep_var[1] = parse_variable (lexer, dict);
856 ll_push_tail (&factor_list, &sf->ll);
859 ll_push_tail (&factor_list, &sf->ll);
861 lex_match (lexer, ',');
863 if ( lex_token (lexer) == '.' || lex_token (lexer) == '/' )
866 success = examine_parse_independent_vars (lexer, dict, cmd);
875 examine_group (struct cmd_examine *cmd, struct casereader *reader, int level,
876 const struct dictionary *dict, struct xfactor *factor)
879 const struct variable *wv = dict_get_weight (dict);
882 struct factor_result *result = xzalloc (sizeof (*result));
885 for (i = 0; i < 2; i++)
886 if (factor->indep_var[i])
887 value_init (&result->value[i], var_get_width (factor->indep_var[i]));
889 result->metrics = xcalloc (n_dependent_vars, sizeof (*result->metrics));
891 if ( cmd->a_statistics[XMN_ST_EXTREME] )
892 n_extrema = cmd->st_n;
895 c = casereader_peek (reader, 0);
899 for (i = 0; i < 2; i++)
900 if (factor->indep_var[i])
901 value_copy (&result->value[i], case_data (c, factor->indep_var[i]),
902 var_get_width (factor->indep_var[i]));
906 for (v = 0; v < n_dependent_vars; ++v)
908 struct casewriter *writer;
909 struct casereader *input = casereader_clone (reader);
911 result->metrics[v].moments = moments1_create (MOMENT_KURTOSIS);
912 result->metrics[v].minima = extrema_create (n_extrema, EXTREME_MINIMA);
913 result->metrics[v].maxima = extrema_create (n_extrema, EXTREME_MAXIMA);
914 result->metrics[v].cmin = DBL_MAX;
916 if (cmd->a_statistics[XMN_ST_DESCRIPTIVES] ||
917 cmd->a_plot[XMN_PLT_BOXPLOT] ||
918 cmd->a_plot[XMN_PLT_NPPLOT] ||
919 cmd->sbc_percentiles)
921 /* In this case, we need to sort the data, so we create a sorting
923 struct subcase up_ordering;
924 subcase_init_var (&up_ordering, dependent_vars[v], SC_ASCEND);
925 writer = sort_create_writer (&up_ordering,
926 casereader_get_proto (reader));
927 subcase_destroy (&up_ordering);
931 /* but in this case, sorting is unnecessary, so an ordinary
932 casewriter is sufficient */
934 autopaging_writer_create (casereader_get_proto (reader));
938 /* Sort or just iterate, whilst calculating moments etc */
939 while ((c = casereader_read (input)) != NULL)
941 int n_vals = caseproto_get_n_widths (casereader_get_proto (reader));
942 const casenumber loc = case_data_idx (c, n_vals - 1)->f;
944 const double weight = wv ? case_data (c, wv)->f : 1.0;
945 const union value *value = case_data (c, dependent_vars[v]);
947 if (weight != SYSMIS)
948 minimize (&result->metrics[v].cmin, weight);
950 moments1_add (result->metrics[v].moments,
954 result->metrics[v].n += weight;
956 if ( ! var_is_value_missing (dependent_vars[v], value, MV_ANY) )
957 result->metrics[v].n_valid += weight;
959 extrema_add (result->metrics[v].maxima,
964 extrema_add (result->metrics[v].minima,
969 casewriter_write (writer, c);
971 casereader_destroy (input);
972 result->metrics[v].up_reader = casewriter_make_reader (writer);
975 /* If percentiles or descriptives have been requested, then a
976 second pass through the data (which has now been sorted)
978 if ( cmd->a_statistics[XMN_ST_DESCRIPTIVES] ||
979 cmd->a_plot[XMN_PLT_BOXPLOT] ||
980 cmd->a_plot[XMN_PLT_NPPLOT] ||
981 cmd->sbc_percentiles)
983 for (v = 0; v < n_dependent_vars; ++v)
987 struct order_stats **os ;
988 struct factor_metrics *metric = &result->metrics[v];
990 metric->n_ptiles = percentile_list.n_data;
992 metric->ptl = xcalloc (metric->n_ptiles,
993 sizeof (struct percentile *));
995 metric->quartiles = xcalloc (3, sizeof (*metric->quartiles));
997 for (i = 0 ; i < metric->n_ptiles; ++i)
999 metric->ptl[i] = (struct percentile *)
1000 percentile_create (percentile_list.data[i] / 100.0, metric->n_valid);
1002 if ( percentile_list.data[i] == 25)
1003 metric->quartiles[0] = metric->ptl[i];
1004 else if ( percentile_list.data[i] == 50)
1005 metric->quartiles[1] = metric->ptl[i];
1006 else if ( percentile_list.data[i] == 75)
1007 metric->quartiles[2] = metric->ptl[i];
1010 metric->tukey_hinges = tukey_hinges_create (metric->n_valid, metric->cmin);
1011 metric->trimmed_mean = trimmed_mean_create (metric->n_valid, 0.05);
1013 n_os = metric->n_ptiles + 2;
1015 if ( cmd->a_plot[XMN_PLT_NPPLOT] )
1017 metric->np = np_create (metric->moments);
1021 os = xcalloc (sizeof (struct order_stats *), n_os);
1023 for (i = 0 ; i < metric->n_ptiles ; ++i )
1025 os[i] = (struct order_stats *) metric->ptl[i];
1028 os[i] = (struct order_stats *) metric->tukey_hinges;
1029 os[i+1] = (struct order_stats *) metric->trimmed_mean;
1031 if (cmd->a_plot[XMN_PLT_NPPLOT])
1032 os[i+2] = metric->np;
1034 order_stats_accumulate (os, n_os,
1035 casereader_clone (metric->up_reader),
1036 wv, dependent_vars[v], MV_ANY);
1041 /* FIXME: Do this in the above loop */
1042 if ( cmd->a_plot[XMN_PLT_HISTOGRAM] )
1045 struct casereader *input = casereader_clone (reader);
1047 for (v = 0; v < n_dependent_vars; ++v)
1049 const struct extremum *max, *min;
1050 struct factor_metrics *metric = &result->metrics[v];
1052 const struct ll_list *max_list =
1053 extrema_list (result->metrics[v].maxima);
1055 const struct ll_list *min_list =
1056 extrema_list (result->metrics[v].minima);
1058 if ( ll_is_empty (max_list))
1060 msg (MW, _("Not creating plot because data set is empty."));
1064 assert (! ll_is_empty (min_list));
1066 max = (const struct extremum *)
1067 ll_data (ll_head(max_list), struct extremum, ll);
1069 min = (const struct extremum *)
1070 ll_data (ll_head (min_list), struct extremum, ll);
1072 metric->histogram = histogram_create (10, min->value, max->value);
1075 while ((c = casereader_read (input)) != NULL)
1077 const double weight = wv ? case_data (c, wv)->f : 1.0;
1079 for (v = 0; v < n_dependent_vars; ++v)
1081 struct factor_metrics *metric = &result->metrics[v];
1082 if ( metric->histogram)
1083 histogram_add ((struct histogram *) metric->histogram,
1084 case_data (c, dependent_vars[v])->f, weight);
1088 casereader_destroy (input);
1091 /* In this case, a third iteration is required */
1092 if (cmd->a_plot[XMN_PLT_BOXPLOT])
1094 for (v = 0; v < n_dependent_vars; ++v)
1096 struct factor_metrics *metric = &result->metrics[v];
1097 int n_vals = caseproto_get_n_widths (casereader_get_proto (
1098 metric->up_reader));
1100 metric->box_whisker =
1101 box_whisker_create ((struct tukey_hinges *) metric->tukey_hinges,
1102 cmd->v_id, n_vals - 1);
1104 order_stats_accumulate ((struct order_stats **) &metric->box_whisker,
1106 casereader_clone (metric->up_reader),
1107 wv, dependent_vars[v], MV_ANY);
1111 ll_push_tail (&factor->result_list, &result->ll);
1112 casereader_destroy (reader);
1117 run_examine (struct cmd_examine *cmd, struct casereader *input,
1121 const struct dictionary *dict = dataset_dict (ds);
1123 struct casereader *level0 = casereader_clone (input);
1125 c = casereader_peek (input, 0);
1128 casereader_destroy (input);
1132 output_split_file_values (ds, c);
1135 ll_init (&level0_factor.result_list);
1137 examine_group (cmd, level0, 0, dict, &level0_factor);
1139 for (ll = ll_head (&factor_list);
1140 ll != ll_null (&factor_list);
1143 struct xfactor *factor = ll_data (ll, struct xfactor, ll);
1145 struct casereader *group = NULL;
1146 struct casereader *level1;
1147 struct casegrouper *grouper1 = NULL;
1149 level1 = casereader_clone (input);
1150 level1 = sort_execute_1var (level1, factor->indep_var[0]);
1151 grouper1 = casegrouper_create_vars (level1, &factor->indep_var[0], 1);
1153 while (casegrouper_get_next_group (grouper1, &group))
1155 struct casereader *group_copy = casereader_clone (group);
1157 if ( !factor->indep_var[1])
1158 examine_group (cmd, group_copy, 1, dict, factor);
1162 struct casereader *group2 = NULL;
1163 struct casegrouper *grouper2 = NULL;
1165 group_copy = sort_execute_1var (group_copy,
1166 factor->indep_var[1]);
1168 grouper2 = casegrouper_create_vars (group_copy,
1169 &factor->indep_var[1], 1);
1171 while (casegrouper_get_next_group (grouper2, &group2))
1173 examine_group (cmd, group2, 2, dict, factor);
1176 casegrouper_destroy (grouper2);
1179 casereader_destroy (group);
1181 casegrouper_destroy (grouper1);
1184 casereader_destroy (input);
1186 output_examine (dict);
1188 factor_destroy (&level0_factor);
1192 for (ll = ll_head (&factor_list);
1193 ll != ll_null (&factor_list);
1196 struct xfactor *f = ll_data (ll, struct xfactor, ll);
1205 show_summary (const struct variable **dependent_var, int n_dep_var,
1206 const struct dictionary *dict,
1207 const struct xfactor *fctr)
1209 const struct variable *wv = dict_get_weight (dict);
1210 const struct fmt_spec *wfmt = wv ? var_get_print_format (wv) : & F_8_0;
1212 static const char *subtitle[]=
1220 int heading_columns = 1;
1222 const int heading_rows = 3;
1223 struct tab_table *tbl;
1230 if ( fctr->indep_var[0] )
1232 heading_columns = 2;
1234 if ( fctr->indep_var[1] )
1236 heading_columns = 3;
1240 n_rows *= ll_count (&fctr->result_list);
1241 n_rows += heading_rows;
1243 n_cols = heading_columns + 6;
1245 tbl = tab_create (n_cols, n_rows, 0);
1246 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
1248 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
1250 /* Outline the box */
1255 n_cols - 1, n_rows - 1);
1257 /* Vertical lines for the data only */
1262 n_cols - 1, n_rows - 1);
1265 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
1266 tab_hline (tbl, TAL_1, heading_columns, n_cols - 1, 1 );
1267 tab_hline (tbl, TAL_1, heading_columns, n_cols - 1, heading_rows -1 );
1269 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
1272 tab_title (tbl, _("Case Processing Summary"));
1274 tab_joint_text (tbl, heading_columns, 0,
1276 TAB_CENTER | TAT_TITLE,
1279 /* Remove lines ... */
1286 for (j = 0 ; j < 3 ; ++j)
1288 tab_text (tbl, heading_columns + j * 2 , 2, TAB_CENTER | TAT_TITLE,
1291 tab_text (tbl, heading_columns + j * 2 + 1, 2, TAB_CENTER | TAT_TITLE,
1294 tab_joint_text (tbl, heading_columns + j * 2 , 1,
1295 heading_columns + j * 2 + 1, 1,
1296 TAB_CENTER | TAT_TITLE,
1299 tab_box (tbl, -1, -1,
1301 heading_columns + j * 2, 1,
1302 heading_columns + j * 2 + 1, 1);
1306 /* Titles for the independent variables */
1307 if ( fctr->indep_var[0] )
1309 tab_text (tbl, 1, heading_rows - 1, TAB_CENTER | TAT_TITLE,
1310 var_to_string (fctr->indep_var[0]));
1312 if ( fctr->indep_var[1] )
1314 tab_text (tbl, 2, heading_rows - 1, TAB_CENTER | TAT_TITLE,
1315 var_to_string (fctr->indep_var[1]));
1319 for (v = 0 ; v < n_dep_var ; ++v)
1323 const union value *last_value = NULL;
1326 tab_hline (tbl, TAL_1, 0, n_cols -1 ,
1327 v * ll_count (&fctr->result_list)
1332 v * ll_count (&fctr->result_list) + heading_rows,
1333 TAB_LEFT | TAT_TITLE,
1334 var_to_string (dependent_var[v])
1338 for (ll = ll_head (&fctr->result_list);
1339 ll != ll_null (&fctr->result_list); ll = ll_next (ll))
1342 const struct factor_result *result =
1343 ll_data (ll, struct factor_result, ll);
1345 if ( fctr->indep_var[0] )
1348 if ( last_value == NULL ||
1349 !value_equal (last_value, &result->value[0],
1350 var_get_width (fctr->indep_var[0])))
1354 last_value = &result->value[0];
1355 ds_init_empty (&str);
1357 var_append_value_name (fctr->indep_var[0], &result->value[0],
1362 v * ll_count (&fctr->result_list),
1363 TAB_LEFT | TAT_TITLE,
1368 if ( fctr->indep_var[1] && j > 0)
1369 tab_hline (tbl, TAL_1, 1, n_cols - 1,
1371 v * ll_count (&fctr->result_list));
1374 if ( fctr->indep_var[1])
1378 ds_init_empty (&str);
1380 var_append_value_name (fctr->indep_var[1],
1381 &result->value[1], &str);
1385 v * ll_count (&fctr->result_list),
1386 TAB_LEFT | TAT_TITLE,
1394 moments1_calculate (result->metrics[v].moments,
1395 &n, &result->metrics[v].mean,
1396 &result->metrics[v].variance,
1397 &result->metrics[v].skewness,
1398 &result->metrics[v].kurtosis);
1400 result->metrics[v].se_mean = sqrt (result->metrics[v].variance / n) ;
1403 tab_double (tbl, heading_columns,
1404 heading_rows + j + v * ll_count (&fctr->result_list),
1408 tab_text (tbl, heading_columns + 1,
1409 heading_rows + j + v * ll_count (&fctr->result_list),
1410 TAB_RIGHT | TAT_PRINTF,
1411 "%g%%", n * 100.0 / result->metrics[v].n);
1414 tab_double (tbl, heading_columns + 2,
1415 heading_rows + j + v * ll_count (&fctr->result_list),
1417 result->metrics[v].n - n,
1420 tab_text (tbl, heading_columns + 3,
1421 heading_rows + j + v * ll_count (&fctr->result_list),
1422 TAB_RIGHT | TAT_PRINTF,
1424 (result->metrics[v].n - n) * 100.0 / result->metrics[v].n
1427 /* Total Valid + Missing */
1428 tab_double (tbl, heading_columns + 4,
1429 heading_rows + j + v * ll_count (&fctr->result_list),
1431 result->metrics[v].n,
1434 tab_text (tbl, heading_columns + 5,
1435 heading_rows + j + v * ll_count (&fctr->result_list),
1436 TAB_RIGHT | TAT_PRINTF,
1438 (result->metrics[v].n) * 100.0 / result->metrics[v].n
1449 #define DESCRIPTIVE_ROWS 13
1452 show_descriptives (const struct variable **dependent_var,
1454 const struct xfactor *fctr)
1457 int heading_columns = 3;
1459 const int heading_rows = 1;
1460 struct tab_table *tbl;
1467 if ( fctr->indep_var[0] )
1469 heading_columns = 4;
1471 if ( fctr->indep_var[1] )
1473 heading_columns = 5;
1477 n_rows *= ll_count (&fctr->result_list) * DESCRIPTIVE_ROWS;
1478 n_rows += heading_rows;
1480 n_cols = heading_columns + 2;
1482 tbl = tab_create (n_cols, n_rows, 0);
1483 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
1485 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
1487 /* Outline the box */
1492 n_cols - 1, n_rows - 1);
1495 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
1496 tab_hline (tbl, TAL_2, 1, n_cols - 1, heading_rows );
1498 tab_vline (tbl, TAL_1, n_cols - 1, 0, n_rows - 1);
1501 if ( fctr->indep_var[0])
1502 tab_text (tbl, 1, 0, TAT_TITLE, var_to_string (fctr->indep_var[0]));
1504 if ( fctr->indep_var[1])
1505 tab_text (tbl, 2, 0, TAT_TITLE, var_to_string (fctr->indep_var[1]));
1507 for (v = 0 ; v < n_dep_var ; ++v )
1512 const int row_var_start =
1513 v * DESCRIPTIVE_ROWS * ll_count(&fctr->result_list);
1517 heading_rows + row_var_start,
1518 TAB_LEFT | TAT_TITLE,
1519 var_to_string (dependent_var[v])
1522 for (ll = ll_head (&fctr->result_list);
1523 ll != ll_null (&fctr->result_list); i++, ll = ll_next (ll))
1525 const struct factor_result *result =
1526 ll_data (ll, struct factor_result, ll);
1529 gsl_cdf_tdist_Qinv ((1 - cmd.n_cinterval[0] / 100.0) / 2.0,
1530 result->metrics[v].n - 1);
1532 if ( i > 0 || v > 0 )
1534 const int left_col = (i == 0) ? 0 : 1;
1535 tab_hline (tbl, TAL_1, left_col, n_cols - 1,
1536 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS);
1539 if ( fctr->indep_var[0])
1542 ds_init_empty (&vstr);
1543 var_append_value_name (fctr->indep_var[0],
1544 &result->value[0], &vstr);
1547 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1556 tab_text (tbl, n_cols - 4,
1557 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1561 tab_text (tbl, n_cols - 4,
1562 heading_rows + row_var_start + 1 + i * DESCRIPTIVE_ROWS,
1563 TAB_LEFT | TAT_PRINTF,
1564 _("%g%% Confidence Interval for Mean"),
1565 cmd.n_cinterval[0]);
1567 tab_text (tbl, n_cols - 3,
1568 heading_rows + row_var_start + 1 + i * DESCRIPTIVE_ROWS,
1572 tab_text (tbl, n_cols - 3,
1573 heading_rows + row_var_start + 2 + i * DESCRIPTIVE_ROWS,
1577 tab_text (tbl, n_cols - 4,
1578 heading_rows + row_var_start + 3 + i * DESCRIPTIVE_ROWS,
1579 TAB_LEFT | TAT_PRINTF,
1580 _("5%% Trimmed Mean"));
1582 tab_text (tbl, n_cols - 4,
1583 heading_rows + row_var_start + 4 + i * DESCRIPTIVE_ROWS,
1587 tab_text (tbl, n_cols - 4,
1588 heading_rows + row_var_start + 5 + i * DESCRIPTIVE_ROWS,
1592 tab_text (tbl, n_cols - 4,
1593 heading_rows + row_var_start + 6 + i * DESCRIPTIVE_ROWS,
1595 _("Std. Deviation"));
1597 tab_text (tbl, n_cols - 4,
1598 heading_rows + row_var_start + 7 + i * DESCRIPTIVE_ROWS,
1602 tab_text (tbl, n_cols - 4,
1603 heading_rows + row_var_start + 8 + i * DESCRIPTIVE_ROWS,
1607 tab_text (tbl, n_cols - 4,
1608 heading_rows + row_var_start + 9 + i * DESCRIPTIVE_ROWS,
1612 tab_text (tbl, n_cols - 4,
1613 heading_rows + row_var_start + 10 + i * DESCRIPTIVE_ROWS,
1615 _("Interquartile Range"));
1618 tab_text (tbl, n_cols - 4,
1619 heading_rows + row_var_start + 11 + i * DESCRIPTIVE_ROWS,
1623 tab_text (tbl, n_cols - 4,
1624 heading_rows + row_var_start + 12 + i * DESCRIPTIVE_ROWS,
1629 /* Now the statistics ... */
1631 tab_double (tbl, n_cols - 2,
1632 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1634 result->metrics[v].mean,
1637 tab_double (tbl, n_cols - 1,
1638 heading_rows + row_var_start + i * DESCRIPTIVE_ROWS,
1640 result->metrics[v].se_mean,
1644 tab_double (tbl, n_cols - 2,
1645 heading_rows + row_var_start + 1 + i * DESCRIPTIVE_ROWS,
1647 result->metrics[v].mean - t *
1648 result->metrics[v].se_mean,
1651 tab_double (tbl, n_cols - 2,
1652 heading_rows + row_var_start + 2 + i * DESCRIPTIVE_ROWS,
1654 result->metrics[v].mean + t *
1655 result->metrics[v].se_mean,
1659 tab_double (tbl, n_cols - 2,
1660 heading_rows + row_var_start + 3 + i * DESCRIPTIVE_ROWS,
1662 trimmed_mean_calculate ((struct trimmed_mean *) result->metrics[v].trimmed_mean),
1666 tab_double (tbl, n_cols - 2,
1667 heading_rows + row_var_start + 4 + i * DESCRIPTIVE_ROWS,
1669 percentile_calculate (result->metrics[v].quartiles[1], percentile_algorithm),
1673 tab_double (tbl, n_cols - 2,
1674 heading_rows + row_var_start + 5 + i * DESCRIPTIVE_ROWS,
1676 result->metrics[v].variance,
1679 tab_double (tbl, n_cols - 2,
1680 heading_rows + row_var_start + 6 + i * DESCRIPTIVE_ROWS,
1682 sqrt (result->metrics[v].variance),
1685 tab_double (tbl, n_cols - 2,
1686 heading_rows + row_var_start + 10 + i * DESCRIPTIVE_ROWS,
1688 percentile_calculate (result->metrics[v].quartiles[2],
1689 percentile_algorithm) -
1690 percentile_calculate (result->metrics[v].quartiles[0],
1691 percentile_algorithm),
1695 tab_double (tbl, n_cols - 2,
1696 heading_rows + row_var_start + 11 + i * DESCRIPTIVE_ROWS,
1698 result->metrics[v].skewness,
1701 tab_double (tbl, n_cols - 2,
1702 heading_rows + row_var_start + 12 + i * DESCRIPTIVE_ROWS,
1704 result->metrics[v].kurtosis,
1707 tab_double (tbl, n_cols - 1,
1708 heading_rows + row_var_start + 11 + i * DESCRIPTIVE_ROWS,
1710 calc_seskew (result->metrics[v].n),
1713 tab_double (tbl, n_cols - 1,
1714 heading_rows + row_var_start + 12 + i * DESCRIPTIVE_ROWS,
1716 calc_sekurt (result->metrics[v].n),
1720 struct extremum *minimum, *maximum ;
1722 struct ll *max_ll = ll_head (extrema_list (result->metrics[v].maxima));
1723 struct ll *min_ll = ll_head (extrema_list (result->metrics[v].minima));
1725 maximum = ll_data (max_ll, struct extremum, ll);
1726 minimum = ll_data (min_ll, struct extremum, ll);
1728 tab_double (tbl, n_cols - 2,
1729 heading_rows + row_var_start + 7 + i * DESCRIPTIVE_ROWS,
1734 tab_double (tbl, n_cols - 2,
1735 heading_rows + row_var_start + 8 + i * DESCRIPTIVE_ROWS,
1740 tab_double (tbl, n_cols - 2,
1741 heading_rows + row_var_start + 9 + i * DESCRIPTIVE_ROWS,
1743 maximum->value - minimum->value,
1749 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
1751 tab_title (tbl, _("Descriptives"));
1753 tab_text (tbl, n_cols - 2, 0, TAB_CENTER | TAT_TITLE,
1756 tab_text (tbl, n_cols - 1, 0, TAB_CENTER | TAT_TITLE,
1765 show_extremes (const struct variable **dependent_var,
1767 const struct xfactor *fctr)
1770 int heading_columns = 3;
1772 const int heading_rows = 1;
1773 struct tab_table *tbl;
1780 if ( fctr->indep_var[0] )
1782 heading_columns = 4;
1784 if ( fctr->indep_var[1] )
1786 heading_columns = 5;
1790 n_rows *= ll_count (&fctr->result_list) * cmd.st_n * 2;
1791 n_rows += heading_rows;
1793 n_cols = heading_columns + 2;
1795 tbl = tab_create (n_cols, n_rows, 0);
1796 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
1798 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
1800 /* Outline the box */
1805 n_cols - 1, n_rows - 1);
1808 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
1809 tab_hline (tbl, TAL_2, 1, n_cols - 1, heading_rows );
1810 tab_vline (tbl, TAL_1, n_cols - 1, 0, n_rows - 1);
1812 if ( fctr->indep_var[0])
1813 tab_text (tbl, 1, 0, TAT_TITLE, var_to_string (fctr->indep_var[0]));
1815 if ( fctr->indep_var[1])
1816 tab_text (tbl, 2, 0, TAT_TITLE, var_to_string (fctr->indep_var[1]));
1818 for (v = 0 ; v < n_dep_var ; ++v )
1822 const int row_var_start = v * cmd.st_n * 2 * ll_count(&fctr->result_list);
1826 heading_rows + row_var_start,
1827 TAB_LEFT | TAT_TITLE,
1828 var_to_string (dependent_var[v])
1831 for (ll = ll_head (&fctr->result_list);
1832 ll != ll_null (&fctr->result_list); i++, ll = ll_next (ll))
1837 const int row_result_start = i * cmd.st_n * 2;
1839 const struct factor_result *result =
1840 ll_data (ll, struct factor_result, ll);
1843 tab_hline (tbl, TAL_1, 1, n_cols - 1,
1844 heading_rows + row_var_start + row_result_start);
1846 tab_hline (tbl, TAL_1, heading_columns - 2, n_cols - 1,
1847 heading_rows + row_var_start + row_result_start + cmd.st_n);
1849 for ( e = 1; e <= cmd.st_n; ++e )
1851 tab_text (tbl, n_cols - 3,
1852 heading_rows + row_var_start + row_result_start + e - 1,
1853 TAB_RIGHT | TAT_PRINTF,
1856 tab_text (tbl, n_cols - 3,
1857 heading_rows + row_var_start + row_result_start + cmd.st_n + e - 1,
1858 TAB_RIGHT | TAT_PRINTF,
1863 min_ll = ll_head (extrema_list (result->metrics[v].minima));
1864 for (e = 0; e < cmd.st_n;)
1866 struct extremum *minimum = ll_data (min_ll, struct extremum, ll);
1867 double weight = minimum->weight;
1869 while (weight-- > 0 && e < cmd.st_n)
1871 tab_double (tbl, n_cols - 1,
1872 heading_rows + row_var_start + row_result_start + cmd.st_n + e,
1878 tab_fixed (tbl, n_cols - 2,
1879 heading_rows + row_var_start +
1880 row_result_start + cmd.st_n + e,
1887 min_ll = ll_next (min_ll);
1891 max_ll = ll_head (extrema_list (result->metrics[v].maxima));
1892 for (e = 0; e < cmd.st_n;)
1894 struct extremum *maximum = ll_data (max_ll, struct extremum, ll);
1895 double weight = maximum->weight;
1897 while (weight-- > 0 && e < cmd.st_n)
1899 tab_double (tbl, n_cols - 1,
1900 heading_rows + row_var_start +
1901 row_result_start + e,
1907 tab_fixed (tbl, n_cols - 2,
1908 heading_rows + row_var_start +
1909 row_result_start + e,
1916 max_ll = ll_next (max_ll);
1920 if ( fctr->indep_var[0])
1923 ds_init_empty (&vstr);
1924 var_append_value_name (fctr->indep_var[0],
1925 &result->value[0], &vstr);
1928 heading_rows + row_var_start + row_result_start,
1937 tab_text (tbl, n_cols - 4,
1938 heading_rows + row_var_start + row_result_start,
1942 tab_text (tbl, n_cols - 4,
1943 heading_rows + row_var_start + row_result_start + cmd.st_n,
1949 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
1952 tab_title (tbl, _("Extreme Values"));
1955 tab_text (tbl, n_cols - 2, 0, TAB_CENTER | TAT_TITLE,
1959 tab_text (tbl, n_cols - 1, 0, TAB_CENTER | TAT_TITLE,
1965 #define PERCENTILE_ROWS 2
1968 show_percentiles (const struct variable **dependent_var,
1970 const struct xfactor *fctr)
1974 int heading_columns = 2;
1976 const int n_percentiles = subc_list_double_count (&percentile_list);
1977 const int heading_rows = 2;
1978 struct tab_table *tbl;
1985 if ( fctr->indep_var[0] )
1987 heading_columns = 3;
1989 if ( fctr->indep_var[1] )
1991 heading_columns = 4;
1995 n_rows *= ll_count (&fctr->result_list) * PERCENTILE_ROWS;
1996 n_rows += heading_rows;
1998 n_cols = heading_columns + n_percentiles;
2000 tbl = tab_create (n_cols, n_rows, 0);
2001 tab_headers (tbl, heading_columns, 0, heading_rows, 0);
2003 tab_dim (tbl, tab_natural_dimensions, NULL, NULL);
2005 /* Outline the box */
2010 n_cols - 1, n_rows - 1);
2013 tab_hline (tbl, TAL_2, 0, n_cols - 1, heading_rows );
2014 tab_hline (tbl, TAL_2, 1, n_cols - 1, heading_rows );
2016 if ( fctr->indep_var[0])
2017 tab_text (tbl, 1, 1, TAT_TITLE, var_to_string (fctr->indep_var[0]));
2019 if ( fctr->indep_var[1])
2020 tab_text (tbl, 2, 1, TAT_TITLE, var_to_string (fctr->indep_var[1]));
2022 for (v = 0 ; v < n_dep_var ; ++v )
2028 const int row_var_start =
2029 v * PERCENTILE_ROWS * ll_count(&fctr->result_list);
2033 heading_rows + row_var_start,
2034 TAB_LEFT | TAT_TITLE,
2035 var_to_string (dependent_var[v])
2038 for (ll = ll_head (&fctr->result_list);
2039 ll != ll_null (&fctr->result_list); i++, ll = ll_next (ll))
2042 const struct factor_result *result =
2043 ll_data (ll, struct factor_result, ll);
2045 if ( i > 0 || v > 0 )
2047 const int left_col = (i == 0) ? 0 : 1;
2048 tab_hline (tbl, TAL_1, left_col, n_cols - 1,
2049 heading_rows + row_var_start + i * PERCENTILE_ROWS);
2052 if ( fctr->indep_var[0])
2055 ds_init_empty (&vstr);
2056 var_append_value_name (fctr->indep_var[0],
2057 &result->value[0], &vstr);
2060 heading_rows + row_var_start + i * PERCENTILE_ROWS,
2069 tab_text (tbl, n_cols - n_percentiles - 1,
2070 heading_rows + row_var_start + i * PERCENTILE_ROWS,
2072 ptile_alg_desc [percentile_algorithm]);
2075 tab_text (tbl, n_cols - n_percentiles - 1,
2076 heading_rows + row_var_start + 1 + i * PERCENTILE_ROWS,
2078 _("Tukey's Hinges"));
2081 tab_vline (tbl, TAL_1, n_cols - n_percentiles -1, heading_rows, n_rows - 1);
2083 tukey_hinges_calculate ((struct tukey_hinges *) result->metrics[v].tukey_hinges,
2086 for (j = 0; j < n_percentiles; ++j)
2088 double hinge = SYSMIS;
2089 tab_double (tbl, n_cols - n_percentiles + j,
2090 heading_rows + row_var_start + i * PERCENTILE_ROWS,
2092 percentile_calculate (result->metrics[v].ptl[j],
2093 percentile_algorithm),
2097 if ( result->metrics[v].ptl[j]->ptile == 0.5)
2099 else if ( result->metrics[v].ptl[j]->ptile == 0.25)
2101 else if ( result->metrics[v].ptl[j]->ptile == 0.75)
2104 if ( hinge != SYSMIS)
2105 tab_double (tbl, n_cols - n_percentiles + j,
2106 heading_rows + row_var_start + 1 + i * PERCENTILE_ROWS,
2116 tab_vline (tbl, TAL_2, heading_columns, 0, n_rows - 1);
2118 tab_title (tbl, _("Percentiles"));
2121 for (i = 0 ; i < n_percentiles; ++i )
2123 tab_text (tbl, n_cols - n_percentiles + i, 1,
2124 TAB_CENTER | TAT_TITLE | TAT_PRINTF,
2126 subc_list_double_at (&percentile_list, i)
2132 tab_joint_text (tbl,
2133 n_cols - n_percentiles, 0,
2135 TAB_CENTER | TAT_TITLE,
2138 /* Vertical lines for the data only */
2142 n_cols - n_percentiles, 1,
2143 n_cols - 1, n_rows - 1);
2145 tab_hline (tbl, TAL_1, n_cols - n_percentiles, n_cols - 1, 1);
2153 factor_to_string_concise (const struct xfactor *fctr,
2154 const struct factor_result *result,
2158 if (fctr->indep_var[0])
2160 var_append_value_name (fctr->indep_var[0], &result->value[0], str);
2162 if ( fctr->indep_var[1] )
2164 ds_put_cstr (str, ",");
2166 var_append_value_name (fctr->indep_var[1], &result->value[1], str);
2168 ds_put_cstr (str, ")");
2175 factor_to_string (const struct xfactor *fctr,
2176 const struct factor_result *result,
2180 if (fctr->indep_var[0])
2182 ds_put_format (str, "(%s = ", var_get_name (fctr->indep_var[0]));
2184 var_append_value_name (fctr->indep_var[0], &result->value[0], str);
2186 if ( fctr->indep_var[1] )
2188 ds_put_cstr (str, ",");
2189 ds_put_format (str, "%s = ", var_get_name (fctr->indep_var[1]));
2191 var_append_value_name (fctr->indep_var[1], &result->value[1], str);
2193 ds_put_cstr (str, ")");