X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=src%2Flanguage%2Fstats%2Ffrequencies.q;h=b5ad8c47793eef15064a887023eead28732c8845;hb=6da9ed01e15953f12bc1ba35fbee53a7a7e46da0;hp=aab345b67ec7fa33dc351020fae9c406f97a3436;hpb=e369cd707a79d3f85d652c59721f42dc7199e614;p=pspp diff --git a/src/language/stats/frequencies.q b/src/language/stats/frequencies.q index aab345b67e..b5ad8c4779 100644 --- a/src/language/stats/frequencies.q +++ b/src/language/stats/frequencies.q @@ -1,5 +1,5 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 1997-9, 2000, 2007, 2009, 2010 Free Software Foundation, Inc. + Copyright (C) 1997-9, 2000, 2007, 2009, 2010, 2011 Free Software Foundation, Inc. This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -23,9 +23,9 @@ #include "data/case.h" #include "data/casegrouper.h" #include "data/casereader.h" +#include "data/dataset.h" #include "data/dictionary.h" #include "data/format.h" -#include "data/procedure.h" #include "data/settings.h" #include "data/value-labels.h" #include "data/variable.h" @@ -43,6 +43,8 @@ #include "libpspp/str.h" #include "math/histogram.h" #include "math/moments.h" +#include "math/chart-geometry.h" + #include "output/chart-item.h" #include "output/charts/piechart.h" #include "output/charts/plot-hist.h" @@ -125,10 +127,6 @@ struct percentile { double p; /* the %ile to be calculated */ double value; /* the %ile's value */ - double x1; /* The datum value <= the percentile */ - double x2; /* The datum value >= the percentile */ - int flag; - int flag2; /* Set to 1 if this percentile value has been found */ bool show; /* True to show this percentile in the statistics box. */ }; @@ -192,7 +190,6 @@ struct var_freqs /* Variable attributes. */ int width; - struct fmt_spec print; }; struct frq_proc @@ -222,8 +219,9 @@ struct frq_proc static void determine_charts (struct frq_proc *, const struct cmd_frequencies *); -static void calc_stats (const struct frq_proc *, const struct var_freqs *, - double d[FRQ_N_STATS]); +static void calc_stats (const struct var_freqs *, double d[FRQ_N_STATS]); +static void calc_percentiles (const struct frq_proc *, + const struct var_freqs *); static void precalc (struct frq_proc *, struct casereader *, struct dataset *); static void calc (struct frq_proc *, const struct ccase *, @@ -497,18 +495,21 @@ postcalc (struct frq_proc *frq, const struct dataset *ds) double d[FRQ_N_STATS]; struct histogram *histogram; - calc_stats (frq, vf, d); + calc_stats (vf, d); histogram = freq_tab_to_hist (frq, &vf->tab, vf->var); - chart_item_submit (histogram_chart_create ( + if ( histogram) + { + chart_item_submit (histogram_chart_create ( histogram->gsl_hist, var_to_string(vf->var), vf->tab.valid_cases, d[FRQ_MEAN], d[FRQ_STDDEV], frq->hist->draw_normal)); - statistic_destroy (&histogram->parent); + statistic_destroy (&histogram->parent); + } } if (frq->pie) @@ -621,10 +622,10 @@ frq_custom_variables (struct lexer *lexer, struct dataset *ds, size_t n_vars; size_t i; - lex_match (lexer, '='); + lex_match (lexer, T_EQUALS); if (lex_token (lexer) != T_ALL && (lex_token (lexer) != T_ID - || dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) == NULL)) + || dict_lookup_var (dataset_dict (ds), lex_tokcstr (lexer)) == NULL)) return 2; /* Get list of current variables, to avoid duplicates. */ @@ -649,7 +650,6 @@ frq_custom_variables (struct lexer *lexer, struct dataset *ds, vf->n_groups = 0; vf->groups = NULL; vf->width = var_get_width (var); - vf->print = *var_get_print_format (var); } frq->n_vars = n_vars; @@ -665,8 +665,9 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc { struct frq_proc *frq = frq_; - lex_match (lexer, '='); - if ((lex_token (lexer) == T_ID && dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) != NULL) + lex_match (lexer, T_EQUALS); + if ((lex_token (lexer) == T_ID + && dict_lookup_var (dataset_dict (ds), lex_tokcstr (lexer)) != NULL) || lex_token (lexer) == T_ID) for (;;) { @@ -683,7 +684,7 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc if (!parse_variables_const (lexer, dataset_dict (ds), &v, &n, PV_NO_DUPLICATE | PV_NUMERIC)) return 0; - if (lex_match (lexer, '(')) + if (lex_match (lexer, T_LPAREN)) { nl = ml = 0; dl = NULL; @@ -696,14 +697,14 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc } dl[nl++] = lex_tokval (lexer); lex_get (lexer); - lex_match (lexer, ','); + lex_match (lexer, T_COMMA); } /* Note that nl might still be 0 and dl might still be NULL. That's okay. */ - if (!lex_match (lexer, ')')) + if (!lex_match (lexer, T_RPAREN)) { free (v); - msg (SE, _("`)' expected after GROUPED interval list.")); + lex_error_expecting (lexer, "`)'", NULL_SENTINEL); return 0; } } @@ -740,14 +741,23 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc } free (v); - if (!lex_match (lexer, '/')) - break; - if ((lex_token (lexer) != T_ID || dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) != NULL) - && lex_token (lexer) != T_ALL) - { - lex_put_back (lexer, '/'); - break; - } + if (lex_token (lexer) != T_SLASH) + break; + + if ((lex_next_token (lexer, 1) == T_ID + && dict_lookup_var (dataset_dict (ds), + lex_next_tokcstr (lexer, 1))) + || lex_next_token (lexer, 1) == T_ALL) + { + /* The token after the slash is a variable name. Keep parsing. */ + lex_get (lexer); + } + else + { + /* The token after the slash must be the start of a new + subcommand. Let the caller see the slash. */ + break; + } } return 1; @@ -843,7 +853,7 @@ dump_freq_table (const struct var_freqs *vf, const struct variable *wv) if (label != NULL) tab_text (t, 0, r, TAB_LEFT, label); - tab_value (t, 1, r, TAB_NONE, &f->value, ft->dict, &vf->print); + tab_value (t, 1, r, TAB_NONE, &f->value, vf->var, NULL); tab_double (t, 2, r, TAB_NONE, f->count, wfmt); tab_double (t, 3, r, TAB_NONE, percent, NULL); tab_double (t, 4, r, TAB_NONE, valid_percent, NULL); @@ -860,7 +870,7 @@ dump_freq_table (const struct var_freqs *vf, const struct variable *wv) if (label != NULL) tab_text (t, 0, r, TAB_LEFT, label); - tab_value (t, 1, r, TAB_NONE, &f->value, ft->dict, &vf->print); + tab_value (t, 1, r, TAB_NONE, &f->value, vf->var, NULL); tab_double (t, 2, r, TAB_NONE, f->count, wfmt); tab_double (t, 3, r, TAB_NONE, f->count / ft->total_cases * 100.0, NULL); @@ -883,114 +893,73 @@ dump_freq_table (const struct var_freqs *vf, const struct variable *wv) /* Statistical display. */ -/* Calculates all the pertinent statistics for variable V, putting them in - array D[]. */ +static double +calc_percentile (double p, double valid_cases, double x1, double x2) +{ + double s, dummy; + + s = (settings_get_algorithm () != COMPATIBLE + ? modf ((valid_cases - 1) * p, &dummy) + : modf ((valid_cases + 1) * p - 1, &dummy)); + + return x1 + (x2 - x1) * s; +} + +/* Calculates all of the percentiles for VF within FRQ. */ static void -calc_stats (const struct frq_proc *frq, - const struct var_freqs *vf, double d[FRQ_N_STATS]) +calc_percentiles (const struct frq_proc *frq, const struct var_freqs *vf) { const struct freq_tab *ft = &vf->tab; double W = ft->valid_cases; - struct moments *m; - struct freq *f=0; - double prev_value; - int most_often; - double X_mode; - + const struct freq *f; + int percentile_idx; double rank; - int i = 0; - int idx; - - /* Calculate percentiles. */ assert (ft->n_valid > 0); - for (i = 0; i < frq->n_percentiles; i++) - { - struct percentile *pc = &frq->percentiles[i]; - - pc->flag = 0; - pc->flag2 = 0; - } - rank = 0; - prev_value = SYSMIS; - for (idx = 0; idx < ft->n_valid; ++idx) + percentile_idx = 0; + for (f = ft->valid; f < ft->missing; f++) { - f = &ft->valid[idx]; - rank += f->count ; - for (i = 0; i < frq->n_percentiles; i++) + rank += f->count; + for (; percentile_idx < frq->n_percentiles; percentile_idx++) { - struct percentile *pc = &frq->percentiles[i]; - double tp; + struct percentile *pc = &frq->percentiles[percentile_idx]; + double tp; - if ( pc->flag2 ) continue ; + tp = (settings_get_algorithm () == ENHANCED + ? (W - 1) * pc->p + : (W + 1) * pc->p - 1); - if ( settings_get_algorithm () != COMPATIBLE ) - tp = - (ft->valid_cases - 1) * pc->p; - else - tp = - (ft->valid_cases + 1) * pc->p - 1; + if (rank <= tp) + break; - if ( pc->flag ) - { - pc->x2 = f->value.f; - pc->x1 = prev_value; - pc->flag2 = 1; - continue; - } - - if (rank > tp ) - { - if ( f->count > 1 && rank - (f->count - 1) > tp ) - { - pc->x2 = pc->x1 = f->value.f; - pc->flag2 = 1; - } - else - { - pc->flag=1; - } - - continue; - } + if (tp + 1 < rank || f + 1 >= ft->missing) + pc->value = f->value.f; + else + pc->value = calc_percentile (pc->p, W, f->value.f, f[1].value.f); } - prev_value = f->value.f; } - - for (i = 0; i < frq->n_percentiles; i++) + for (; percentile_idx < frq->n_percentiles; percentile_idx++) { - struct percentile *pc = &frq->percentiles[i]; - - /* Catches the case when p == 100% */ - if ( ! pc->flag2 ) - pc->x1 = pc->x2 = f->value.f; - - /* - printf("percentile %d (p==%.2f); X1 = %g; X2 = %g\n", - i,pc->p,pc->x1,pc->x2); - */ + struct percentile *pc = &frq->percentiles[percentile_idx]; + pc->value = ft->valid[ft->n_valid - 1].value.f; } +} - for (i = 0; i < frq->n_percentiles; i++) - { - struct percentile *pc = &frq->percentiles[i]; - double s; - - double dummy; - if ( settings_get_algorithm () != COMPATIBLE ) - { - s = modf((ft->valid_cases - 1) * pc->p , &dummy); - } - else - { - s = modf((ft->valid_cases + 1) * pc->p -1, &dummy); - } - - pc->value = pc->x1 + (pc->x2 - pc->x1) * s ; - } +/* Calculates all the pertinent statistics for VF, putting them in array + D[]. */ +static void +calc_stats (const struct var_freqs *vf, double d[FRQ_N_STATS]) +{ + const struct freq_tab *ft = &vf->tab; + double W = ft->valid_cases; + const struct freq *f; + struct moments *m; + int most_often; + double X_mode; + assert (ft->n_valid > 0); /* Calculate the mode. */ most_often = -1; @@ -1052,7 +1021,8 @@ dump_statistics (const struct frq_proc *frq, const struct var_freqs *vf, var_get_name (vf->var)); return; } - calc_stats (frq, vf, stat_value); + calc_stats (vf, stat_value); + calc_percentiles (frq, vf); t = tab_create (3, frq->n_stats + frq->n_show_percentiles + 2); @@ -1080,7 +1050,7 @@ dump_statistics (const struct frq_proc *frq, const struct var_freqs *vf, tab_double (t, 2, 0, TAB_NONE, ft->valid_cases, wfmt); tab_double (t, 2, 1, TAB_NONE, ft->total_cases - ft->valid_cases, wfmt); - for (i = 0; i < frq->n_percentiles; i++, r++) + for (i = 0; i < frq->n_percentiles; i++) { struct percentile *pc = &frq->percentiles[i]; @@ -1098,6 +1068,7 @@ dump_statistics (const struct frq_proc *frq, const struct var_freqs *vf, tab_fixed (t, 1, r, TAB_LEFT, pc->p * 100, 3, 0); tab_double (t, 2, r, TAB_NONE, pc->value, var_get_print_format (vf->var)); + r++; } tab_title (t, "%s", var_to_string (vf->var)); @@ -1148,10 +1119,9 @@ freq_tab_to_hist (const struct frq_proc *frq, const struct freq_tab *ft, { double x_min, x_max, valid_freq; int i; - + double bin_width; struct histogram *histogram; double iqr; - int bins; /* Find out the extremes of the x value, within the range to be included in the histogram, and sum the total frequency of those values. */ @@ -1171,19 +1141,13 @@ freq_tab_to_hist (const struct frq_proc *frq, const struct freq_tab *ft, /* Freedman-Diaconis' choice of bin width. */ iqr = calculate_iqr (frq); - if (iqr != SYSMIS) - { - double bin_width = 2 * iqr / pow (valid_freq, 1.0 / 3.0); - bins = (x_max - x_min) / bin_width; - if (bins < 5) - bins = 5; - else if (bins > 400) - bins = 400; - } - else - bins = 5; + bin_width = 2 * iqr / pow (valid_freq, 1.0 / 3.0); + + histogram = histogram_create (bin_width, x_min, x_max); + + if ( histogram == NULL) + return NULL; - histogram = histogram_create (bins, x_min, x_max); for (i = 0; i < ft->n_valid; i++) { const struct freq *f = &ft->valid[i];