1 /* PSPP - A program for statistical analysis . -*-c-*-
3 Copyright (C) 2004 Free Software Foundation, Inc.
4 Author: John Darrington 2004
6 This program is free software; you can redistribute it and/or
7 modify it under the terms of the GNU General Public License as
8 published by the Free Software Foundation; either version 2 of the
9 License, or (at your option) any later version.
11 This program is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
21 #include "factor_stats.h"
25 #include "algorithm.h"
36 metrics_precalc(struct metrics *fs)
47 fs->ordered_data = hsh_create(20,
48 (hsh_compare_func *) compare_values,
49 (hsh_hash_func *) hash_value,
50 (hsh_free_func *) weighted_value_free,
56 /* Include val in the calculation for the metrics.
57 If val is null, then treat it as MISSING
60 metrics_calc(struct metrics *fs, const union value *val,
61 double weight, int case_no)
63 struct weighted_value **wv;
68 fs->n_missing += weight;
74 fs->ssq += x * x * weight;
75 fs->sum += x * weight;
77 if ( x < fs->min) fs->min = x;
78 if ( x > fs->max) fs->max = x;
81 wv = (struct weighted_value **) hsh_probe (fs->ordered_data,(void *) val );
85 /* If this value has already been seen, then simply
86 increase its weight and push a new case number */
90 assert( (*wv)->v.f == val->f );
93 cn = xmalloc( sizeof (struct case_node) ) ;
94 cn->next = (*wv)->case_nos ;
101 struct case_node *cn;
103 *wv = weighted_value_create();
107 cn = xmalloc( sizeof (struct case_node) ) ;
110 (*wv)->case_nos = cn;
117 metrics_postcalc(struct metrics *m)
126 struct weighted_value **data;
131 m->mean = m->sum / m->n;
133 sample_var = ( m->ssq / m->n - m->mean * m->mean );
135 m->var = m->n * sample_var / ( m->n - 1) ;
136 m->stddev = sqrt(m->var);
139 /* FIXME: Check this is correct ???
140 Shouldn't we use the sample variance ??? */
141 m->stderr = sqrt (m->var / m->n) ;
143 data = (struct weighted_value **) hsh_data(m->ordered_data);
144 n_data = hsh_count(m->ordered_data);
148 m->trimmed_mean = m->mean;
153 m->wv = xmalloc(sizeof(struct weighted_value ) * n_data);
155 for ( i = 0 ; i < n_data ; ++i )
156 m->wv[i] = *(data[i]);
158 sort (m->wv, n_data, sizeof (struct weighted_value) ,
159 (algo_compare_func *) compare_values, 0);
162 /* Trimmed mean calculation */
169 for ( i = 0 ; i < n_data ; ++i )
174 m->wv[i].rank = j + (m->wv[i].w - 1) / 2.0 ;
184 for ( i = n_data -1 ; i >= 0; --i )
186 if ( tc > m->n - m->wv[i].cc)
192 for ( i = k1 + 2 ; i <= k2 - 1 ; ++i )
194 m->trimmed_mean += m->wv[i].v.f * m->wv[i].w;
198 m->trimmed_mean += (m->n - m->wv[k2 - 1].cc - tc) * m->wv[k2].v.f ;
199 m->trimmed_mean += (m->wv[k1 + 1].cc - tc) * m->wv[k1 + 1].v.f ;
200 m->trimmed_mean /= 0.9 * m->n ;
205 struct weighted_value *
206 weighted_value_create(void)
208 struct weighted_value *wv;
209 wv = xmalloc (sizeof (struct weighted_value ));
218 weighted_value_free(struct weighted_value *wv)
220 struct case_node *cn = wv->case_nos;
224 struct case_node *next = cn->next;
238 /* Create a factor statistics object with for N dependent vars
239 and ID as the value of the independent variable */
240 struct factor_statistics *
241 create_factor_statistics (int n, union value *id0, union value *id1)
243 struct factor_statistics *f;
245 f = xmalloc( sizeof ( struct factor_statistics ));
249 f->m = xmalloc( sizeof ( struct metrics ) * n ) ;
256 factor_statistics_free(struct factor_statistics *f)
269 factor_statistics_compare(const struct factor_statistics *f0,
270 const struct factor_statistics *f1, void *aux)
278 cmp0 = compare_values(&f0->id[0], &f1->id[0], aux);
284 if ( ( f0->id[1].f == SYSMIS ) && (f1->id[1].f != SYSMIS) )
287 if ( ( f0->id[1].f != SYSMIS ) && (f1->id[1].f == SYSMIS) )
290 return compare_values(&f0->id[1], &f1->id[1], aux);
295 factor_statistics_hash(const struct factor_statistics *f, void *aux)
300 h = hash_value(&f->id[0], aux);
302 if ( f->id[1].f != SYSMIS )
303 h += hash_value(&f->id[1], aux);