1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 2005, 2009, 2010, 2011, 2012, 2013, 2014,
3 2016, 2017, 2019 Free Software Foundation, Inc.
5 This program is free software: you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation, either version 3 of the License, or
8 (at your option) any later version.
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program. If not, see <http://www.gnu.org/licenses/>. */
23 #include <gsl/gsl_math.h>
24 #include <gsl/gsl_cdf.h>
25 #include <gsl/gsl_matrix.h>
27 #include <data/dataset.h>
28 #include <data/casewriter.h>
30 #include "language/command.h"
31 #include "language/lexer/lexer.h"
32 #include "language/lexer/value-parser.h"
33 #include "language/lexer/variable-parser.h"
36 #include "data/casegrouper.h"
37 #include "data/casereader.h"
38 #include "data/dictionary.h"
40 #include "math/covariance.h"
41 #include "math/linreg.h"
42 #include "math/moments.h"
44 #include "libpspp/message.h"
45 #include "libpspp/taint.h"
47 #include "output/pivot-table.h"
49 #include "gl/intprops.h"
50 #include "gl/minmax.h"
53 #define _(msgid) gettext (msgid)
54 #define N_(msgid) msgid
65 #define STATS_DEFAULT (STATS_R | STATS_COEFF | STATS_ANOVA | STATS_OUTS)
73 const struct variable **vars;
76 const struct variable **dep_vars;
88 struct regression_workspace
90 /* The new variables which will be introduced by /SAVE */
91 const struct variable **predvars;
92 const struct variable **residvars;
94 /* A reader/writer pair to temporarily hold the
95 values of the new variables */
96 struct casewriter *writer;
97 struct casereader *reader;
99 /* Indeces of the new values in the reader/writer (-1 if not applicable) */
103 /* 0, 1 or 2 depending on what new variables are to be created */
107 static void run_regression (const struct regression *cmd,
108 struct regression_workspace *ws,
109 struct casereader *input);
112 /* Return a string based on PREFIX which may be used as the name
113 of a new variable in DICT */
115 reg_get_name (const struct dictionary *dict, const char *prefix)
120 /* XXX handle too-long prefixes */
121 name = xmalloc (strlen (prefix) + INT_BUFSIZE_BOUND (i) + 1);
124 sprintf (name, "%s%d", prefix, i);
125 if (dict_lookup_var (dict, name) == NULL)
131 static const struct variable *
132 create_aux_var (struct dataset *ds, const char *prefix)
134 struct variable *var;
135 struct dictionary *dict = dataset_dict (ds);
136 char *name = reg_get_name (dict, prefix);
137 var = dict_create_var_assert (dict, name, 0);
142 /* Auxiliary data for transformation when /SAVE is entered */
143 struct save_trans_data
146 struct regression_workspace *ws;
150 save_trans_free (void *aux)
152 struct save_trans_data *save_trans_data = aux;
153 free (save_trans_data->ws->predvars);
154 free (save_trans_data->ws->residvars);
156 casereader_destroy (save_trans_data->ws->reader);
157 free (save_trans_data->ws);
158 free (save_trans_data);
162 static enum trns_result
163 save_trans_func (void *aux, struct ccase **c, casenumber x UNUSED)
165 struct save_trans_data *save_trans_data = aux;
166 struct regression_workspace *ws = save_trans_data->ws;
167 struct ccase *in = casereader_read (ws->reader);
172 *c = case_unshare (*c);
174 for (k = 0; k < save_trans_data->n_dep_vars; ++k)
176 if (ws->pred_idx != -1)
178 double pred = case_num_idx (in, ws->extras * k + ws->pred_idx);
179 *case_num_rw (*c, ws->predvars[k]) = pred;
182 if (ws->res_idx != -1)
184 double resid = case_num_idx (in, ws->extras * k + ws->res_idx);
185 *case_num_rw (*c, ws->residvars[k]) = resid;
191 return TRNS_CONTINUE;
195 cmd_regression (struct lexer *lexer, struct dataset *ds)
197 struct regression_workspace workspace;
198 struct regression regression;
199 const struct dictionary *dict = dataset_dict (ds);
202 memset (®ression, 0, sizeof (struct regression));
204 regression.ci = 0.95;
205 regression.stats = STATS_DEFAULT;
206 regression.pred = false;
207 regression.resid = false;
210 regression.origin = false;
212 bool variables_seen = false;
213 bool method_seen = false;
214 bool dependent_seen = false;
215 while (lex_token (lexer) != T_ENDCMD)
217 lex_match (lexer, T_SLASH);
219 if (lex_match_id (lexer, "VARIABLES"))
223 msg (SE, _("VARIABLES may not appear after %s"), "METHOD");
228 msg (SE, _("VARIABLES may not appear after %s"), "DEPENDENT");
231 variables_seen = true;
232 lex_match (lexer, T_EQUALS);
234 if (!parse_variables_const (lexer, dict,
235 ®ression.vars, ®ression.n_vars,
236 PV_NO_DUPLICATE | PV_NUMERIC))
239 else if (lex_match_id (lexer, "DEPENDENT"))
241 dependent_seen = true;
242 lex_match (lexer, T_EQUALS);
244 free (regression.dep_vars);
245 regression.n_dep_vars = 0;
247 if (!parse_variables_const (lexer, dict,
248 ®ression.dep_vars,
249 ®ression.n_dep_vars,
250 PV_NO_DUPLICATE | PV_NUMERIC))
253 else if (lex_match_id (lexer, "ORIGIN"))
255 regression.origin = true;
257 else if (lex_match_id (lexer, "NOORIGIN"))
259 regression.origin = false;
261 else if (lex_match_id (lexer, "METHOD"))
264 lex_match (lexer, T_EQUALS);
266 if (!lex_force_match_id (lexer, "ENTER"))
271 if (! variables_seen)
273 if (!parse_variables_const (lexer, dict,
274 ®ression.vars, ®ression.n_vars,
275 PV_NO_DUPLICATE | PV_NUMERIC))
279 else if (lex_match_id (lexer, "STATISTICS"))
281 unsigned long statistics = 0;
282 lex_match (lexer, T_EQUALS);
284 while (lex_token (lexer) != T_ENDCMD
285 && lex_token (lexer) != T_SLASH)
287 if (lex_match (lexer, T_ALL))
291 else if (lex_match_id (lexer, "DEFAULTS"))
293 statistics |= STATS_DEFAULT;
295 else if (lex_match_id (lexer, "R"))
297 statistics |= STATS_R;
299 else if (lex_match_id (lexer, "COEFF"))
301 statistics |= STATS_COEFF;
303 else if (lex_match_id (lexer, "ANOVA"))
305 statistics |= STATS_ANOVA;
307 else if (lex_match_id (lexer, "BCOV"))
309 statistics |= STATS_BCOV;
311 else if (lex_match_id (lexer, "TOL"))
313 statistics |= STATS_TOL;
315 else if (lex_match_id (lexer, "CI"))
317 statistics |= STATS_CI;
319 if (lex_match (lexer, T_LPAREN) &&
320 lex_force_num (lexer))
322 regression.ci = lex_number (lexer) / 100.0;
324 if (! lex_force_match (lexer, T_RPAREN))
330 lex_error (lexer, NULL);
336 regression.stats = statistics;
339 else if (lex_match_id (lexer, "SAVE"))
341 lex_match (lexer, T_EQUALS);
343 while (lex_token (lexer) != T_ENDCMD
344 && lex_token (lexer) != T_SLASH)
346 if (lex_match_id (lexer, "PRED"))
348 regression.pred = true;
350 else if (lex_match_id (lexer, "RESID"))
352 regression.resid = true;
356 lex_error (lexer, NULL);
363 lex_error (lexer, NULL);
368 if (!regression.vars)
370 dict_get_vars (dict, ®ression.vars, ®ression.n_vars, 0);
373 save = regression.pred || regression.resid;
374 workspace.extras = 0;
375 workspace.res_idx = -1;
376 workspace.pred_idx = -1;
377 workspace.writer = NULL;
378 workspace.reader = NULL;
379 workspace.residvars = NULL;
380 workspace.predvars = NULL;
384 struct caseproto *proto = caseproto_create ();
386 if (regression.resid)
388 workspace.res_idx = workspace.extras ++;
389 workspace.residvars = xcalloc (regression.n_dep_vars, sizeof (*workspace.residvars));
391 for (i = 0; i < regression.n_dep_vars; ++i)
393 workspace.residvars[i] = create_aux_var (ds, "RES");
394 proto = caseproto_add_width (proto, 0);
400 workspace.pred_idx = workspace.extras ++;
401 workspace.predvars = xcalloc (regression.n_dep_vars, sizeof (*workspace.predvars));
403 for (i = 0; i < regression.n_dep_vars; ++i)
405 workspace.predvars[i] = create_aux_var (ds, "PRED");
406 proto = caseproto_add_width (proto, 0);
410 if (proc_make_temporary_transformations_permanent (ds))
411 msg (SW, _("REGRESSION with SAVE ignores TEMPORARY. "
412 "Temporary transformations will be made permanent."));
414 if (dict_get_filter (dict))
415 msg (SW, _("REGRESSION with SAVE ignores FILTER. "
416 "All cases will be processed."));
418 workspace.writer = autopaging_writer_create (proto);
419 caseproto_unref (proto);
424 struct casegrouper *grouper;
425 struct casereader *group;
428 grouper = casegrouper_create_splits (proc_open_filtering (ds, !save), dict);
431 while (casegrouper_get_next_group (grouper, &group))
433 run_regression (®ression,
438 ok = casegrouper_destroy (grouper);
439 ok = proc_commit (ds) && ok;
442 if (workspace.writer)
444 struct save_trans_data *save_trans_data = xmalloc (sizeof *save_trans_data);
445 struct casereader *r = casewriter_make_reader (workspace.writer);
446 workspace.writer = NULL;
447 workspace.reader = r;
448 save_trans_data->ws = xmalloc (sizeof (workspace));
449 memcpy (save_trans_data->ws, &workspace, sizeof (workspace));
450 save_trans_data->n_dep_vars = regression.n_dep_vars;
452 static const struct trns_class trns_class = {
453 .name = "REGRESSION",
454 .execute = save_trans_func,
455 .destroy = save_trans_free,
457 add_transformation (ds, &trns_class, save_trans_data);
461 free (regression.vars);
462 free (regression.dep_vars);
467 free (regression.vars);
468 free (regression.dep_vars);
472 /* Return the size of the union of dependent and independent variables */
474 get_n_all_vars (const struct regression *cmd)
476 size_t result = cmd->n_vars;
480 result += cmd->n_dep_vars;
481 for (i = 0; i < cmd->n_dep_vars; i++)
483 for (j = 0; j < cmd->n_vars; j++)
485 if (cmd->vars[j] == cmd->dep_vars[i])
494 /* Fill VARS with the union of dependent and independent variables */
496 fill_all_vars (const struct variable **vars, const struct regression *cmd)
500 for (i = 0; i < cmd->n_vars; i++)
502 vars[i] = cmd->vars[i];
505 for (i = 0; i < cmd->n_dep_vars; i++)
509 for (j = 0; j < cmd->n_vars; j++)
511 if (cmd->dep_vars[i] == cmd->vars[j])
519 vars[cmd->n_vars + x++] = cmd->dep_vars[i];
525 /* Fill the array VARS, with all the predictor variables from CMD, except
528 fill_predictor_x (const struct variable **vars, const struct variable *x, const struct regression *cmd)
533 for (i = 0; i < cmd->n_vars; i++)
535 if (cmd->vars[i] == x)
538 vars[n++] = cmd->vars[i];
543 Is variable k the dependent variable?
546 is_depvar (const struct regression *cmd, size_t k, const struct variable *v)
548 return v == cmd->vars[k];
552 /* Identify the explanatory variables in v_variables. Returns
553 the number of independent variables. */
555 identify_indep_vars (const struct regression *cmd,
556 const struct variable **indep_vars,
557 const struct variable *depvar)
559 int n_indep_vars = 0;
562 for (i = 0; i < cmd->n_vars; i++)
563 if (!is_depvar (cmd, i, depvar))
564 indep_vars[n_indep_vars++] = cmd->vars[i];
565 if ((n_indep_vars < 1) && is_depvar (cmd, 0, depvar))
568 There is only one independent variable, and it is the same
569 as the dependent variable. Print a warning and continue.
573 ("The dependent variable is equal to the independent variable. "
574 "The least squares line is therefore Y=X. "
575 "Standard errors and related statistics may be meaningless."));
577 indep_vars[0] = cmd->vars[0];
583 fill_covariance (gsl_matrix * cov, struct covariance *all_cov,
584 const struct variable **vars,
585 size_t n_vars, const struct variable *dep_var,
586 const struct variable **all_vars, size_t n_all_vars,
591 size_t dep_subscript = SIZE_MAX;
593 const gsl_matrix *ssizes;
594 const gsl_matrix *mean_matrix;
595 const gsl_matrix *ssize_matrix;
598 const gsl_matrix *cm = covariance_calculate_unnormalized (all_cov);
603 rows = xnmalloc (cov->size1 - 1, sizeof (*rows));
605 for (i = 0; i < n_all_vars; i++)
607 for (j = 0; j < n_vars; j++)
609 if (vars[j] == all_vars[i])
614 if (all_vars[i] == dep_var)
619 assert (dep_subscript != SIZE_MAX);
621 mean_matrix = covariance_moments (all_cov, MOMENT_MEAN);
622 ssize_matrix = covariance_moments (all_cov, MOMENT_NONE);
623 for (i = 0; i < cov->size1 - 1; i++)
625 means[i] = gsl_matrix_get (mean_matrix, rows[i], 0)
626 / gsl_matrix_get (ssize_matrix, rows[i], 0);
627 for (j = 0; j < cov->size2 - 1; j++)
629 gsl_matrix_set (cov, i, j, gsl_matrix_get (cm, rows[i], rows[j]));
630 gsl_matrix_set (cov, j, i, gsl_matrix_get (cm, rows[j], rows[i]));
633 means[cov->size1 - 1] = gsl_matrix_get (mean_matrix, dep_subscript, 0)
634 / gsl_matrix_get (ssize_matrix, dep_subscript, 0);
635 ssizes = covariance_moments (all_cov, MOMENT_NONE);
636 result = gsl_matrix_get (ssizes, dep_subscript, rows[0]);
637 for (i = 0; i < cov->size1 - 1; i++)
639 gsl_matrix_set (cov, i, cov->size1 - 1,
640 gsl_matrix_get (cm, rows[i], dep_subscript));
641 gsl_matrix_set (cov, cov->size1 - 1, i,
642 gsl_matrix_get (cm, rows[i], dep_subscript));
643 if (result > gsl_matrix_get (ssizes, rows[i], dep_subscript))
645 result = gsl_matrix_get (ssizes, rows[i], dep_subscript);
648 gsl_matrix_set (cov, cov->size1 - 1, cov->size1 - 1,
649 gsl_matrix_get (cm, dep_subscript, dep_subscript));
656 struct model_container
658 struct linreg **models;
662 STATISTICS subcommand output functions.
664 static void reg_stats_r (const struct linreg *, const struct variable *);
665 static void reg_stats_coeff (const struct regression *, const struct linreg *,
666 const struct model_container *, const gsl_matrix *,
667 const struct variable *);
668 static void reg_stats_anova (const struct linreg *, const struct variable *);
669 static void reg_stats_bcov (const struct linreg *, const struct variable *);
672 static struct linreg **
673 run_regression_get_models (const struct regression *cmd,
674 struct casereader *input,
678 struct model_container *model_container = XCALLOC (cmd->n_vars, struct model_container);
681 struct covariance *cov;
682 struct casereader *reader;
684 if (cmd->stats & STATS_TOL)
686 for (i = 0; i < cmd->n_vars; i++)
688 struct regression subreg;
689 subreg.origin = cmd->origin;
691 subreg.n_vars = cmd->n_vars - 1;
692 subreg.n_dep_vars = 1;
693 subreg.vars = xmalloc (sizeof (*subreg.vars) * cmd->n_vars - 1);
694 subreg.dep_vars = xmalloc (sizeof (*subreg.dep_vars));
695 fill_predictor_x (subreg.vars, cmd->vars[i], cmd);
696 subreg.dep_vars[0] = cmd->vars[i];
697 subreg.stats = STATS_R;
699 subreg.resid = false;
702 model_container[i].models =
703 run_regression_get_models (&subreg, input, false);
705 free (subreg.dep_vars);
709 size_t n_all_vars = get_n_all_vars (cmd);
710 const struct variable **all_vars = xnmalloc (n_all_vars, sizeof (*all_vars));
712 /* In the (rather pointless) case where the dependent variable is
713 the independent variable, n_all_vars == 1.
714 However this would result in a buffer overflow so we must
715 over-allocate the space required in this malloc call.
717 double *means = xnmalloc (n_all_vars <= 1 ? 2 : n_all_vars,
719 fill_all_vars (all_vars, cmd);
720 cov = covariance_1pass_create (n_all_vars, all_vars,
721 dict_get_weight (dataset_dict (cmd->ds)),
722 MV_ANY, cmd->origin == false);
724 reader = casereader_clone (input);
725 reader = casereader_create_filter_missing (reader, all_vars, n_all_vars,
728 struct casereader *r = casereader_clone (reader);
730 for (; (c = casereader_read (r)) != NULL; case_unref (c))
732 covariance_accumulate (cov, c);
734 casereader_destroy (r);
737 struct linreg **models = XCALLOC (cmd->n_dep_vars, struct linreg*);
739 for (int k = 0; k < cmd->n_dep_vars; k++)
741 const struct variable **vars = xnmalloc (cmd->n_vars, sizeof (*vars));
742 const struct variable *dep_var = cmd->dep_vars[k];
743 int n_indep = identify_indep_vars (cmd, vars, dep_var);
744 gsl_matrix *cov_matrix = gsl_matrix_alloc (n_indep + 1, n_indep + 1);
745 double n_data = fill_covariance (cov_matrix, cov, vars, n_indep,
746 dep_var, all_vars, n_all_vars, means);
747 models[k] = linreg_alloc (dep_var, vars, n_data, n_indep, cmd->origin);
748 for (i = 0; i < n_indep; i++)
750 linreg_set_indep_variable_mean (models[k], i, means[i]);
752 linreg_set_depvar_mean (models[k], means[i]);
755 linreg_fit (cov_matrix, models[k]);
757 if (output && !taint_has_tainted_successor (casereader_get_taint (input)))
760 Find the least-squares estimates and other statistics.
762 if (cmd->stats & STATS_R)
763 reg_stats_r (models[k], dep_var);
765 if (cmd->stats & STATS_ANOVA)
766 reg_stats_anova (models[k], dep_var);
768 if (cmd->stats & STATS_COEFF)
769 reg_stats_coeff (cmd, models[k],
771 cov_matrix, dep_var);
773 if (cmd->stats & STATS_BCOV)
774 reg_stats_bcov (models[k], dep_var);
779 msg (SE, _("No valid data found. This command was skipped."));
782 gsl_matrix_free (cov_matrix);
785 casereader_destroy (reader);
787 for (int i = 0; i < cmd->n_vars; i++)
789 if (model_container[i].models)
791 linreg_unref (model_container[i].models[0]);
793 free (model_container[i].models);
795 free (model_container);
799 covariance_destroy (cov);
804 run_regression (const struct regression *cmd,
805 struct regression_workspace *ws,
806 struct casereader *input)
808 struct linreg **models = run_regression_get_models (cmd, input, true);
813 struct casereader *r = casereader_clone (input);
815 for (; (c = casereader_read (r)) != NULL; case_unref (c))
817 struct ccase *outc = case_create (casewriter_get_proto (ws->writer));
818 for (int k = 0; k < cmd->n_dep_vars; k++)
820 const struct variable **vars = xnmalloc (cmd->n_vars, sizeof (*vars));
821 const struct variable *dep_var = cmd->dep_vars[k];
822 int n_indep = identify_indep_vars (cmd, vars, dep_var);
823 double *vals = xnmalloc (n_indep, sizeof (*vals));
824 for (int i = 0; i < n_indep; i++)
826 const union value *tmp = case_data (c, vars[i]);
832 double pred = linreg_predict (models[k], vals, n_indep);
833 *case_num_rw_idx (outc, k * ws->extras + ws->pred_idx) = pred;
838 double obs = case_num (c, linreg_dep_var (models[k]));
839 double res = linreg_residual (models[k], obs, vals, n_indep);
840 *case_num_rw_idx (outc, k * ws->extras + ws->res_idx) = res;
845 casewriter_write (ws->writer, outc);
847 casereader_destroy (r);
850 for (int k = 0; k < cmd->n_dep_vars; k++)
852 linreg_unref (models[k]);
856 casereader_destroy (input);
863 reg_stats_r (const struct linreg * c, const struct variable *var)
865 struct pivot_table *table = pivot_table_create__ (
866 pivot_value_new_text_format (N_("Model Summary (%s)"),
867 var_to_string (var)),
870 pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Statistics"),
871 N_("R"), N_("R Square"), N_("Adjusted R Square"),
872 N_("Std. Error of the Estimate"));
874 double rsq = linreg_ssreg (c) / linreg_sst (c);
875 double adjrsq = (rsq -
876 (1.0 - rsq) * linreg_n_coeffs (c)
877 / (linreg_n_obs (c) - linreg_n_coeffs (c) - 1));
878 double std_error = sqrt (linreg_mse (c));
881 sqrt (rsq), rsq, adjrsq, std_error
883 for (size_t i = 0; i < sizeof entries / sizeof *entries; i++)
884 pivot_table_put1 (table, i, pivot_value_new_number (entries[i]));
886 pivot_table_submit (table);
890 Table showing estimated regression coefficients.
893 reg_stats_coeff (const struct regression *cmd, const struct linreg *c,
894 const struct model_container *mc, const gsl_matrix *cov,
895 const struct variable *var)
897 struct pivot_table *table = pivot_table_create__ (
898 pivot_value_new_text_format (N_("Coefficients (%s)"), var_to_string (var)),
901 struct pivot_dimension *statistics = pivot_dimension_create (
902 table, PIVOT_AXIS_COLUMN, N_("Statistics"));
903 pivot_category_create_group (statistics->root,
904 N_("Unstandardized Coefficients"),
905 N_("B"), N_("Std. Error"));
906 pivot_category_create_group (statistics->root,
907 N_("Standardized Coefficients"), N_("Beta"));
908 pivot_category_create_leaves (statistics->root, N_("t"),
909 N_("Sig."), PIVOT_RC_SIGNIFICANCE);
910 if (cmd->stats & STATS_CI)
912 struct pivot_category *interval = pivot_category_create_group__ (
913 statistics->root, pivot_value_new_text_format (
914 N_("%g%% Confidence Interval for B"),
916 pivot_category_create_leaves (interval, N_("Lower Bound"),
920 if (cmd->stats & STATS_TOL)
921 pivot_category_create_group (statistics->root,
922 N_("Collinearity Statistics"),
923 N_("Tolerance"), N_("VIF"));
926 struct pivot_dimension *variables = pivot_dimension_create (
927 table, PIVOT_AXIS_ROW, N_("Variables"));
929 double df = linreg_n_obs (c) - linreg_n_coeffs (c) - 1;
930 double q = (1 - cmd->ci) / 2.0; /* 2-tailed test */
931 double tval = gsl_cdf_tdist_Qinv (q, df);
935 int var_idx = pivot_category_create_leaf (
936 variables->root, pivot_value_new_text (N_("(Constant)")));
938 double std_err = sqrt (gsl_matrix_get (linreg_cov (c), 0, 0));
939 double t_stat = linreg_intercept (c) / std_err;
940 double base_entries[] = {
941 linreg_intercept (c),
945 2.0 * gsl_cdf_tdist_Q (fabs (t_stat),
946 linreg_n_obs (c) - linreg_n_coeffs (c)),
950 for (size_t i = 0; i < sizeof base_entries / sizeof *base_entries; i++)
951 pivot_table_put2 (table, col++, var_idx,
952 pivot_value_new_number (base_entries[i]));
954 if (cmd->stats & STATS_CI)
956 double interval_entries[] = {
957 linreg_intercept (c) - tval * std_err,
958 linreg_intercept (c) + tval * std_err,
961 for (size_t i = 0; i < sizeof interval_entries / sizeof *interval_entries; i++)
962 pivot_table_put2 (table, col++, var_idx,
963 pivot_value_new_number (interval_entries[i]));
967 for (size_t j = 0; j < linreg_n_coeffs (c); j++)
969 const struct variable *v = linreg_indep_var (c, j);
970 int var_idx = pivot_category_create_leaf (
971 variables->root, pivot_value_new_variable (v));
973 double std_err = sqrt (gsl_matrix_get (linreg_cov (c), j + 1, j + 1));
974 double t_stat = linreg_coeff (c, j) / std_err;
975 double base_entries[] = {
977 sqrt (gsl_matrix_get (linreg_cov (c), j + 1, j + 1)),
978 (sqrt (gsl_matrix_get (cov, j, j)) * linreg_coeff (c, j) /
979 sqrt (gsl_matrix_get (cov, cov->size1 - 1, cov->size2 - 1))),
981 2 * gsl_cdf_tdist_Q (fabs (t_stat), df)
985 for (size_t i = 0; i < sizeof base_entries / sizeof *base_entries; i++)
986 pivot_table_put2 (table, col++, var_idx,
987 pivot_value_new_number (base_entries[i]));
989 if (cmd->stats & STATS_CI)
991 double interval_entries[] = {
992 linreg_coeff (c, j) - tval * std_err,
993 linreg_coeff (c, j) + tval * std_err,
997 for (size_t i = 0; i < sizeof interval_entries / sizeof *interval_entries; i++)
998 pivot_table_put2 (table, col++, var_idx,
999 pivot_value_new_number (interval_entries[i]));
1002 if (cmd->stats & STATS_TOL)
1005 struct linreg *m = mc[j].models[0];
1006 double rsq = linreg_ssreg (m) / linreg_sst (m);
1007 pivot_table_put2 (table, col++, var_idx, pivot_value_new_number (1.0 - rsq));
1008 pivot_table_put2 (table, col++, var_idx, pivot_value_new_number (1.0 / (1.0 - rsq)));
1013 pivot_table_submit (table);
1017 Display the ANOVA table.
1020 reg_stats_anova (const struct linreg * c, const struct variable *var)
1022 struct pivot_table *table = pivot_table_create__ (
1023 pivot_value_new_text_format (N_("ANOVA (%s)"), var_to_string (var)),
1026 pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Statistics"),
1027 N_("Sum of Squares"), PIVOT_RC_OTHER,
1028 N_("df"), PIVOT_RC_INTEGER,
1029 N_("Mean Square"), PIVOT_RC_OTHER,
1030 N_("F"), PIVOT_RC_OTHER,
1031 N_("Sig."), PIVOT_RC_SIGNIFICANCE);
1033 pivot_dimension_create (table, PIVOT_AXIS_ROW, N_("Source"),
1034 N_("Regression"), N_("Residual"), N_("Total"));
1036 double msm = linreg_ssreg (c) / linreg_dfmodel (c);
1037 double mse = linreg_mse (c);
1038 double F = msm / mse;
1047 /* Sums of Squares. */
1048 { 0, 0, linreg_ssreg (c) },
1049 { 0, 1, linreg_sse (c) },
1050 { 0, 2, linreg_sst (c) },
1051 /* Degrees of freedom. */
1052 { 1, 0, linreg_dfmodel (c) },
1053 { 1, 1, linreg_dferror (c) },
1054 { 1, 2, linreg_dftotal (c) },
1061 { 4, 0, gsl_cdf_fdist_Q (F, linreg_dfmodel (c), linreg_dferror (c)) },
1063 for (size_t i = 0; i < sizeof entries / sizeof *entries; i++)
1065 const struct entry *e = &entries[i];
1066 pivot_table_put2 (table, e->stat_idx, e->source_idx,
1067 pivot_value_new_number (e->x));
1070 pivot_table_submit (table);
1075 reg_stats_bcov (const struct linreg * c, const struct variable *var)
1077 struct pivot_table *table = pivot_table_create__ (
1078 pivot_value_new_text_format (N_("Coefficient Correlations (%s)"),
1079 var_to_string (var)),
1080 "Coefficient Correlations");
1082 for (size_t i = 0; i < 2; i++)
1084 struct pivot_dimension *models = pivot_dimension_create (
1085 table, i ? PIVOT_AXIS_ROW : PIVOT_AXIS_COLUMN, N_("Models"));
1086 for (size_t j = 0; j < linreg_n_coeffs (c); j++)
1087 pivot_category_create_leaf (
1088 models->root, pivot_value_new_variable (
1089 linreg_indep_var (c, j)));
1092 pivot_dimension_create (table, PIVOT_AXIS_ROW, N_("Statistics"),
1095 for (size_t i = 0; i < linreg_n_coeffs (c); i++)
1096 for (size_t k = 0; k < linreg_n_coeffs (c); k++)
1098 double cov = gsl_matrix_get (linreg_cov (c), MIN (i, k), MAX (i, k));
1099 pivot_table_put3 (table, k, i, 0, pivot_value_new_number (cov));
1102 pivot_table_submit (table);