X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=src%2Flanguage%2Fstats%2Ffrequencies.q;h=e397eb8760361fe70fa1a679f6955cc9585daebf;hb=740218f508466f16a79939cd97027b99f589d682;hp=aab345b67ec7fa33dc351020fae9c406f97a3436;hpb=e369cd707a79d3f85d652c59721f42dc7199e614;p=pspp diff --git a/src/language/stats/frequencies.q b/src/language/stats/frequencies.q index aab345b67e..e397eb8760 100644 --- a/src/language/stats/frequencies.q +++ b/src/language/stats/frequencies.q @@ -125,10 +125,6 @@ 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. */ }; @@ -222,8 +218,9 @@ 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 *, @@ -497,7 +494,7 @@ 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); @@ -621,7 +618,7 @@ 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)) @@ -665,7 +662,7 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc { struct frq_proc *frq = frq_; - lex_match (lexer, '='); + lex_match (lexer, T_EQUALS); if ((lex_token (lexer) == T_ID && dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) != NULL) || lex_token (lexer) == T_ID) for (;;) @@ -683,7 +680,7 @@ 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; @@ -696,11 +693,11 @@ 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.")); @@ -740,12 +737,12 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc } free (v); - if (!lex_match (lexer, '/')) + if (!lex_match (lexer, T_SLASH)) 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, '/'); + lex_put_back (lexer, T_SLASH); break; } } @@ -883,114 +880,74 @@ 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; - - if ( pc->flag2 ) continue ; - - if ( settings_get_algorithm () != COMPATIBLE ) - tp = - (ft->valid_cases - 1) * pc->p; - else - tp = - (ft->valid_cases + 1) * pc->p - 1; - - 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; - } + struct percentile *pc = &frq->percentiles[percentile_idx]; + double tp; - continue; - } - } - prev_value = f->value.f; - } + tp = (settings_get_algorithm () == ENHANCED + ? (W - 1) * pc->p + : (W + 1) * pc->p - 1); - for (i = 0; i < frq->n_percentiles; i++) - { - struct percentile *pc = &frq->percentiles[i]; - - /* Catches the case when p == 100% */ - if ( ! pc->flag2 ) - pc->x1 = pc->x2 = f->value.f; + if (rank <= tp) + break; - /* - printf("percentile %d (p==%.2f); X1 = %g; X2 = %g\n", - i,pc->p,pc->x1,pc->x2); - */ + if (f->count > 1 + && (rank - (f->count - 1) > tp || f + 1 >= ft->missing)) + pc->value = f->value.f; + else + pc->value = calc_percentile (pc->p, W, f->value.f, f[1].value.f); + } } - - for (i = 0; i < frq->n_percentiles; i++) + for (; percentile_idx < frq->n_percentiles; percentile_idx++) { - 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 ; + struct percentile *pc = &frq->percentiles[percentile_idx]; + pc->value = ft->valid[ft->n_valid - 1].value.f; } +} + +/* 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 +1009,8 @@ 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);