1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 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 <libpspp/assertion.h>
20 #include "covariance.h"
21 #include <gl/xalloc.h>
23 #include <gsl/gsl_matrix.h>
24 #include <data/case.h>
25 #include <data/variable.h>
26 #include <libpspp/misc.h>
27 #include "categoricals.h"
29 #define n_MOMENTS (MOMENT_VARIANCE + 1)
32 /* Create a new matrix of NEW_SIZE x NEW_SIZE and copy the elements of
33 matrix IN into it. IN must be a square matrix, and in normal usage
34 it will be smaller than NEW_SIZE.
35 IN is destroyed by this function. The return value must be destroyed
36 when no longer required.
39 resize_matrix (gsl_matrix *in, size_t new_size)
43 gsl_matrix *out = NULL;
45 assert (in->size1 == in->size2);
47 if (new_size <= in->size1)
50 out = gsl_matrix_calloc (new_size, new_size);
52 for (i = 0; i < in->size1; ++i)
54 for (j = 0; j < in->size2; ++j)
56 double x = gsl_matrix_get (in, i, j);
58 gsl_matrix_set (out, i, j, x);
69 /* The variables for which the covariance matrix is to be calculated. */
71 const struct variable **vars;
73 /* Categorical variables. */
74 struct categoricals *categoricals;
76 /* Array containing number of categories per categorical variable. */
79 /* Dimension of the covariance matrix. */
82 /* The weight variable (or NULL if none) */
83 const struct variable *wv;
85 /* A set of matrices containing the 0th, 1st and 2nd moments */
88 /* The class of missing values to exclude */
89 enum mv_class exclude;
91 /* An array of doubles representing the covariance matrix.
92 Only the top triangle is included, and no diagonals */
96 /* 1 for single pass algorithm;
97 2 for double pass algorithm
102 0 : No pass has been made
103 1 : First pass has been started
104 2 : Second pass has been
106 IE: How many passes have been (partially) made. */
109 /* Flags indicating that the first case has been seen */
110 bool pass_one_first_case_seen;
111 bool pass_two_first_case_seen;
116 /* Return a matrix containing the M th moments.
117 The matrix is of size NxN where N is the number of variables.
118 Each row represents the moments of a variable.
119 In the absence of missing values, the columns of this matrix will
120 be identical. If missing values are involved, then element (i,j)
121 is the moment of the i th variable, when paired with the j th variable.
124 covariance_moments (const struct covariance *cov, int m)
126 return cov->moments[m];
131 /* Create a covariance struct.
134 covariance_1pass_create (size_t n_vars, const struct variable **vars,
135 const struct variable *weight, enum mv_class exclude)
138 struct covariance *cov = xmalloc (sizeof *cov);
142 cov->pass_one_first_case_seen = cov->pass_two_first_case_seen = false;
147 cov->n_vars = n_vars;
150 cov->moments = xmalloc (sizeof *cov->moments * n_MOMENTS);
152 for (i = 0; i < n_MOMENTS; ++i)
153 cov->moments[i] = gsl_matrix_calloc (n_vars, n_vars);
155 cov->exclude = exclude;
157 cov->n_cm = (n_vars * (n_vars - 1) ) / 2;
159 cov->cm = xcalloc (sizeof *cov->cm, cov->n_cm);
165 Create a covariance struct for a two-pass algorithm. If categorical
166 variables are involed, the dimension cannot be know until after the
167 first data pass, so the actual covariances will not be allocated
171 covariance_2pass_create (size_t n_vars, const struct variable **vars,
172 size_t n_catvars, const struct variable **catvars,
173 const struct variable *wv, enum mv_class exclude)
176 struct covariance *cov = xmalloc (sizeof *cov);
180 cov->pass_one_first_case_seen = cov->pass_two_first_case_seen = false;
185 cov->n_vars = n_vars;
188 cov->moments = xmalloc (sizeof *cov->moments * n_MOMENTS);
190 for (i = 0; i < n_MOMENTS; ++i)
191 cov->moments[i] = gsl_matrix_calloc (n_vars, n_vars);
193 cov->exclude = exclude;
198 cov->categoricals = categoricals_create (catvars, n_catvars, wv, exclude);
203 /* Return an integer, which can be used to index
204 into COV->cm, to obtain the I, J th element
205 of the covariance matrix. If COV->cm does not
206 contain that element, then a negative value
210 cm_idx (const struct covariance *cov, int i, int j)
213 const int n2j = cov->dim - 2 - j;
214 const int nj = cov->dim - 2 ;
217 assert (j < cov->dim);
222 if (j >= cov->dim - 1)
229 as -= n2j * (n2j + 1) ;
237 Returns true iff the variable corresponding to the Ith element of the covariance matrix
238 has a missing value for case C
241 is_missing (const struct covariance *cov, int i, const struct ccase *c)
243 const struct variable *var = i < cov->n_vars ?
245 categoricals_get_variable_by_subscript (cov->categoricals, i - cov->n_vars);
247 const union value *val = case_data (c, var);
249 return var_is_value_missing (var, val, cov->exclude);
254 get_val (const struct covariance *cov, int i, const struct ccase *c)
256 if ( i < cov->n_vars)
258 const struct variable *var = cov->vars[i];
260 const union value *val = case_data (c, var);
265 return categoricals_get_binary_by_subscript (cov->categoricals, i - cov->n_vars, c);
269 dump_matrix (const gsl_matrix *m)
273 for (i = 0 ; i < m->size1; ++i)
275 for (j = 0 ; j < m->size2; ++j)
276 printf ("%02f ", gsl_matrix_get (m, i, j));
281 /* Call this function for every case in the data set */
283 covariance_accumulate_pass1 (struct covariance *cov, const struct ccase *c)
286 const double weight = cov->wv ? case_data (c, cov->wv)->f : 1.0;
288 assert (cov->passes == 2);
289 if (!cov->pass_one_first_case_seen)
291 assert (cov->state == 0);
295 categoricals_update (cov->categoricals, c);
297 for (i = 0 ; i < cov->dim; ++i)
299 double v1 = get_val (cov, i, c);
301 if ( is_missing (cov, i, c))
304 for (j = 0 ; j < cov->dim; ++j)
308 if ( is_missing (cov, j, c))
311 for (m = 0 ; m <= MOMENT_MEAN; ++m)
313 double *x = gsl_matrix_ptr (cov->moments[m], i, j);
321 cov->pass_one_first_case_seen = true;
325 /* Call this function for every case in the data set */
327 covariance_accumulate_pass2 (struct covariance *cov, const struct ccase *c)
330 const double weight = cov->wv ? case_data (c, cov->wv)->f : 1.0;
332 assert (cov->passes == 2);
333 assert (cov->state >= 1);
335 if (! cov->pass_two_first_case_seen)
338 assert (cov->state == 1);
341 cov->dim = cov->n_vars + categoricals_total (cov->categoricals);
342 cov->n_cm = (cov->dim * (cov->dim - 1) ) / 2;
343 cov->cm = xcalloc (sizeof *cov->cm, cov->n_cm);
345 /* Grow the moment matrices so that they're large enough to accommodate the
346 categorical elements */
347 for (i = 0; i < n_MOMENTS; ++i)
349 cov->moments[i] = resize_matrix (cov->moments[i], cov->dim);
352 categoricals_done (cov->categoricals);
354 /* Populate the moments matrices with the categorical value elements */
355 for (i = cov->n_vars; i < cov->dim; ++i)
357 for (j = 0 ; j < cov->dim ; ++j) /* FIXME: This is WRONG !!! */
359 double w = categoricals_get_weight_by_subscript (cov->categoricals, i - cov->n_vars);
361 gsl_matrix_set (cov->moments[MOMENT_NONE], i, j, w);
363 w = categoricals_get_sum_by_subscript (cov->categoricals, i - cov->n_vars);
365 gsl_matrix_set (cov->moments[MOMENT_MEAN], i, j, w);
369 /* FIXME: This is WRONG!! It must be fixed to properly handle missing values. For
370 now it assumes there are none */
371 for (m = 0 ; m < n_MOMENTS; ++m)
373 for (i = 0 ; i < cov->dim ; ++i)
375 double x = gsl_matrix_get (cov->moments[m], i, cov->n_vars -1);
376 for (j = cov->n_vars; j < cov->dim; ++j)
378 gsl_matrix_set (cov->moments[m], i, j, x);
383 /* Divide the means by the number of samples */
384 for (i = 0; i < cov->dim; ++i)
386 for (j = 0; j < cov->dim; ++j)
388 double *x = gsl_matrix_ptr (cov->moments[MOMENT_MEAN], i, j);
389 *x /= gsl_matrix_get (cov->moments[MOMENT_NONE], i, j);
394 for (i = 0 ; i < cov->dim; ++i)
396 double v1 = get_val (cov, i, c);
398 if ( is_missing (cov, i, c))
401 for (j = 0 ; j < cov->dim; ++j)
405 double v2 = get_val (cov, j, c);
407 const double s = pow2 (v1 - gsl_matrix_get (cov->moments[MOMENT_MEAN], i, j)) * weight;
409 if ( is_missing (cov, j, c))
413 double *x = gsl_matrix_ptr (cov->moments[MOMENT_VARIANCE], i, j);
418 (v1 - gsl_matrix_get (cov->moments[MOMENT_MEAN], i, j))
420 (v2 - gsl_matrix_get (cov->moments[MOMENT_MEAN], i, j))
424 idx = cm_idx (cov, i, j);
432 cov->pass_two_first_case_seen = true;
436 /* Call this function for every case in the data set.
437 After all cases have been passed, call covariance_calculate
440 covariance_accumulate (struct covariance *cov, const struct ccase *c)
443 const double weight = cov->wv ? case_data (c, cov->wv)->f : 1.0;
445 assert (cov->passes == 1);
447 if ( !cov->pass_one_first_case_seen)
449 assert ( cov->state == 0);
453 for (i = 0 ; i < cov->dim; ++i)
455 const union value *val1 = case_data (c, cov->vars[i]);
457 if ( is_missing (cov, i, c))
460 for (j = 0 ; j < cov->dim; ++j)
464 const union value *val2 = case_data (c, cov->vars[j]);
466 if ( is_missing (cov, j, c))
469 idx = cm_idx (cov, i, j);
472 cov->cm [idx] += val1->f * val2->f * weight;
475 for (m = 0 ; m < n_MOMENTS; ++m)
477 double *x = gsl_matrix_ptr (cov->moments[m], i, j);
485 cov->pass_one_first_case_seen = true;
490 Allocate and return a gsl_matrix containing the covariances of the
494 cm_to_gsl (struct covariance *cov)
497 gsl_matrix *m = gsl_matrix_calloc (cov->dim, cov->dim);
499 /* Copy the non-diagonal elements from cov->cm */
500 for ( j = 0 ; j < cov->dim - 1; ++j)
502 for (i = j+1 ; i < cov->dim; ++i)
504 double x = cov->cm [cm_idx (cov, i, j)];
505 gsl_matrix_set (m, i, j, x);
506 gsl_matrix_set (m, j, i, x);
510 /* Copy the diagonal elements from cov->moments[2] */
511 for (j = 0 ; j < cov->dim ; ++j)
513 double sigma = gsl_matrix_get (cov->moments[2], j, j);
514 gsl_matrix_set (m, j, j, sigma);
521 static const gsl_matrix *
522 covariance_calculate_double_pass (struct covariance *cov)
525 for (i = 0 ; i < cov->dim; ++i)
527 for (j = 0 ; j < cov->dim; ++j)
530 double *x = gsl_matrix_ptr (cov->moments[MOMENT_VARIANCE], i, j);
531 *x /= gsl_matrix_get (cov->moments[MOMENT_NONE], i, j);
533 idx = cm_idx (cov, i, j);
537 *x /= gsl_matrix_get (cov->moments[MOMENT_NONE], i, j);
542 return cm_to_gsl (cov);
545 static const gsl_matrix *
546 covariance_calculate_single_pass (struct covariance *cov)
551 for (m = 0; m < n_MOMENTS; ++m)
553 /* Divide the moments by the number of samples */
556 for (i = 0 ; i < cov->dim; ++i)
558 for (j = 0 ; j < cov->dim; ++j)
560 double *x = gsl_matrix_ptr (cov->moments[m], i, j);
561 *x /= gsl_matrix_get (cov->moments[0], i, j);
563 if ( m == MOMENT_VARIANCE)
564 *x -= pow2 (gsl_matrix_get (cov->moments[1], i, j));
570 /* Centre the moments */
571 for ( j = 0 ; j < cov->dim - 1; ++j)
573 for (i = j + 1 ; i < cov->dim; ++i)
575 double *x = &cov->cm [cm_idx (cov, i, j)];
577 *x /= gsl_matrix_get (cov->moments[0], i, j);
580 gsl_matrix_get (cov->moments[MOMENT_MEAN], i, j)
582 gsl_matrix_get (cov->moments[MOMENT_MEAN], j, i);
586 return cm_to_gsl (cov);
592 Return a pointer to gsl_matrix containing the pairwise covariances.
593 The matrix remains owned by the COV object, and must not be freed.
594 Call this function only after all data have been accumulated.
597 covariance_calculate (struct covariance *cov)
599 assert ( cov->state > 0 );
604 return covariance_calculate_single_pass (cov);
607 return covariance_calculate_double_pass (cov);
617 /* Destroy the COV object */
619 covariance_destroy (struct covariance *cov)
623 categoricals_destroy (cov->categoricals);
625 for (i = 0; i < n_MOMENTS; ++i)
626 gsl_matrix_free (cov->moments[i]);