X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Foneway.c;h=c9ae255ad7508160e59d69d1a55e23fb070b5571;hb=0d323042aa7503f67186c88a5d41d1517ba9926c;hp=ce4f016d99063acabf0dd71abab6d36245229aa3;hpb=27efd5d4d3b65514a11ae4d5afacafbf70373310;p=pspp diff --git a/src/language/stats/oneway.c b/src/language/stats/oneway.c index ce4f016d99..c9ae255ad7 100644 --- a/src/language/stats/oneway.c +++ b/src/language/stats/oneway.c @@ -20,6 +20,11 @@ #include #include +#include +#include +#include +#include + #include #include @@ -81,8 +86,6 @@ struct oneway_spec size_t n_vars; const struct variable **vars; - const struct dictionary *dict; - const struct variable *indep_var; enum statistics stats; @@ -92,6 +95,24 @@ struct oneway_spec /* List of contrasts */ struct ll_list contrast_list; + + /* The weight variable */ + const struct variable *wv; +}; + + +/* Workspace variable for each dependent variable */ +struct per_var_ws +{ + struct covariance *cov; + + double sst; + double sse; + double ssa; + + int n_groups; + + double cc; }; struct oneway_workspace @@ -103,11 +124,13 @@ struct oneway_workspace /* A hash table containing all the distinct values of the independent variable */ struct hsh_table *group_hash; + + struct per_var_ws *vws; }; /* Routines to show the output tables */ -static void show_anova_table (const struct oneway_spec *); -static void show_descriptives (const struct oneway_spec *, const struct dictionary *dict); +static void show_anova_table (const struct oneway_spec *, const struct oneway_workspace *); +static void show_descriptives (const struct oneway_spec *); static void show_homogeneity (const struct oneway_spec *); static void output_oneway (const struct oneway_spec *, struct oneway_workspace *ws); @@ -116,6 +139,7 @@ static void run_oneway (const struct oneway_spec *cmd, struct casereader *input, int cmd_oneway (struct lexer *lexer, struct dataset *ds) { + const struct dictionary *dict = dataset_dict (ds); struct oneway_spec oneway ; oneway.n_vars = 0; oneway.vars = NULL; @@ -123,7 +147,7 @@ cmd_oneway (struct lexer *lexer, struct dataset *ds) oneway.stats = 0; oneway.missing_type = MISS_ANALYSIS; oneway.exclude = MV_ANY; - oneway.dict = dataset_dict (ds); + oneway.wv = dict_get_weight (dict); ll_init (&oneway.contrast_list); @@ -137,14 +161,14 @@ cmd_oneway (struct lexer *lexer, struct dataset *ds) lex_match (lexer, '='); } - if (!parse_variables_const (lexer, oneway.dict, + if (!parse_variables_const (lexer, dict, &oneway.vars, &oneway.n_vars, PV_NO_DUPLICATE | PV_NUMERIC)) goto error; lex_force_match (lexer, T_BY); - oneway.indep_var = parse_variable_const (lexer, oneway.dict); + oneway.indep_var = parse_variable_const (lexer, dict); while (lex_token (lexer) != '.') { @@ -239,7 +263,9 @@ cmd_oneway (struct lexer *lexer, struct dataset *ds) struct casereader *group; bool ok; - grouper = casegrouper_create_splits (proc_open (ds), oneway.dict); + + + grouper = casegrouper_create_splits (proc_open (ds), dict); while (casegrouper_get_next_group (grouper, &group)) run_oneway (&oneway, group, ds); ok = casegrouper_destroy (grouper); @@ -288,13 +314,25 @@ run_oneway (const struct oneway_spec *cmd, struct casereader *input, const struct dataset *ds) { + int v; struct taint *taint; struct dictionary *dict = dataset_dict (ds); struct casereader *reader; struct ccase *c; + struct oneway_workspace ws; + ws.vws = xmalloc (cmd->n_vars * sizeof (*ws.vws)); + + for (v = 0; v < cmd->n_vars; ++v) + { + ws.vws[v].cov = covariance_2pass_create (1, &cmd->vars[v], + 1, &cmd->indep_var, + cmd->wv, cmd->exclude); + ws.vws[v].cc = 0; + } + c = casereader_peek (input, 0); if (c == NULL) { @@ -322,6 +360,7 @@ run_oneway (const struct oneway_spec *cmd, input = casereader_create_filter_weight (input, dict, NULL, NULL); reader = casereader_clone (input); + for (; (c = casereader_read (reader)) != NULL; case_unref (c)) { size_t i; @@ -338,6 +377,13 @@ run_oneway (const struct oneway_spec *cmd, for (i = 0; i < cmd->n_vars; ++i) { + { + struct per_var_ws *pvw = &ws.vws[i]; + + pvw->cc += weight; + covariance_accumulate_pass1 (pvw->cov, c); + } + const struct variable *v = cmd->vars[i]; const union value *val = case_data (c, v); @@ -393,6 +439,34 @@ run_oneway (const struct oneway_spec *cmd, } casereader_destroy (reader); + reader = casereader_clone (input); + for ( ; (c = casereader_read (reader) ); case_unref (c)) + { + int i; + for (i = 0; i < cmd->n_vars; ++i) + { + struct per_var_ws *pvw = &ws.vws[i]; + covariance_accumulate_pass2 (pvw->cov, c); + } + } + casereader_destroy (reader); + + for (v = 0; v < cmd->n_vars; ++v) + { + struct per_var_ws *pvw = &ws.vws[v]; + gsl_matrix *cm = covariance_calculate_unnormalized (pvw->cov); + const struct categoricals *cats = covariance_get_categoricals (pvw->cov); + + pvw->sst = gsl_matrix_get (cm, 0, 0); + + reg_sweep (cm, 0); + + pvw->sse = gsl_matrix_get (cm, 0, 0); + + pvw->ssa = pvw->sst - pvw->sse; + + pvw->n_groups = categoricals_total (cats); + } postcalc (cmd); @@ -512,12 +586,12 @@ output_oneway (const struct oneway_spec *cmd, struct oneway_workspace *ws) } if (cmd->stats & STATS_DESCRIPTIVES) - show_descriptives (cmd, cmd->dict); + show_descriptives (cmd); if (cmd->stats & STATS_HOMOGENEITY) show_homogeneity (cmd); - show_anova_table (cmd); + show_anova_table (cmd, ws); if (ll_count (&cmd->contrast_list) > 0) @@ -541,7 +615,7 @@ output_oneway (const struct oneway_spec *cmd, struct oneway_workspace *ws) /* Show the ANOVA table */ static void -show_anova_table (const struct oneway_spec *cmd) +show_anova_table (const struct oneway_spec *cmd, const struct oneway_workspace *ws) { size_t i; int n_cols =7; @@ -570,21 +644,13 @@ show_anova_table (const struct oneway_spec *cmd) for (i = 0; i < cmd->n_vars; ++i) { - struct group_statistics *totals = &group_proc_get (cmd->vars[i])->ugs; - struct hsh_table *group_hash = group_proc_get (cmd->vars[i])->group_hash; - struct hsh_iterator g; - struct group_statistics *gs; - double ssa = 0; - const char *s = var_to_string (cmd->vars[i]); - - for (gs = hsh_first (group_hash, &g); - gs != 0; - gs = hsh_next (group_hash, &g)) - { - ssa += pow2 (gs->sum) / gs->n; - } + const struct per_var_ws *pvw = &ws->vws[i]; + struct group_proc *gp = group_proc_get (cmd->vars[i]); + const double df1 = pvw->n_groups - 1; + const double df2 = pvw->cc - pvw->n_groups; + const double msa = pvw->ssa / df1; - ssa -= pow2 (totals->sum) / totals->n; + const char *s = var_to_string (cmd->vars[i]); tab_text (t, 0, i * 3 + 1, TAB_LEFT | TAT_TITLE, s); tab_text (t, 1, i * 3 + 1, TAB_LEFT | TAT_TITLE, _("Between Groups")); @@ -594,44 +660,35 @@ show_anova_table (const struct oneway_spec *cmd) if (i > 0) tab_hline (t, TAL_1, 0, n_cols - 1, i * 3 + 1); - { - struct group_proc *gp = group_proc_get (cmd->vars[i]); - const double sst = totals->ssq - pow2 (totals->sum) / totals->n; - const double df1 = gp->n_groups - 1; - const double df2 = totals->n - gp->n_groups; - const double msa = ssa / df1; - - gp->mse = (sst - ssa) / df2; + gp->mse = (pvw->sst - pvw->ssa) / df2; - /* Sums of Squares */ - tab_double (t, 2, i * 3 + 1, 0, ssa, NULL); - tab_double (t, 2, i * 3 + 3, 0, sst, NULL); - tab_double (t, 2, i * 3 + 2, 0, sst - ssa, NULL); + /* Sums of Squares */ + tab_double (t, 2, i * 3 + 1, 0, pvw->ssa, NULL); + tab_double (t, 2, i * 3 + 3, 0, pvw->sst, NULL); + tab_double (t, 2, i * 3 + 2, 0, pvw->sse, NULL); - /* Degrees of freedom */ - tab_fixed (t, 3, i * 3 + 1, 0, df1, 4, 0); - tab_fixed (t, 3, i * 3 + 2, 0, df2, 4, 0); - tab_fixed (t, 3, i * 3 + 3, 0, totals->n - 1, 4, 0); + /* Degrees of freedom */ + tab_fixed (t, 3, i * 3 + 1, 0, df1, 4, 0); + tab_fixed (t, 3, i * 3 + 2, 0, df2, 4, 0); + tab_fixed (t, 3, i * 3 + 3, 0, pvw->cc - 1, 4, 0); - /* Mean Squares */ - tab_double (t, 4, i * 3 + 1, TAB_RIGHT, msa, NULL); - tab_double (t, 4, i * 3 + 2, TAB_RIGHT, gp->mse, NULL); + /* Mean Squares */ + tab_double (t, 4, i * 3 + 1, TAB_RIGHT, msa, NULL); + tab_double (t, 4, i * 3 + 2, TAB_RIGHT, gp->mse, NULL); - { - const double F = msa / gp->mse ; + { + const double F = msa / gp->mse ; - /* The F value */ - tab_double (t, 5, i * 3 + 1, 0, F, NULL); + /* The F value */ + tab_double (t, 5, i * 3 + 1, 0, F, NULL); - /* The significance */ - tab_double (t, 6, i * 3 + 1, 0, gsl_cdf_fdist_Q (F, df1, df2), NULL); - } + /* The significance */ + tab_double (t, 6, i * 3 + 1, 0, gsl_cdf_fdist_Q (F, df1, df2), NULL); } } - tab_title (t, _("ANOVA")); tab_submit (t); } @@ -639,7 +696,7 @@ show_anova_table (const struct oneway_spec *cmd) /* Show the descriptives table */ static void -show_descriptives (const struct oneway_spec *cmd, const struct dictionary *dict) +show_descriptives (const struct oneway_spec *cmd) { size_t v; int n_cols = 10; @@ -649,8 +706,7 @@ show_descriptives (const struct oneway_spec *cmd, const struct dictionary *dict) const double confidence = 0.95; const double q = (1.0 - confidence) / 2.0; - const struct variable *wv = dict_get_weight (dict); - const struct fmt_spec *wfmt = wv ? var_get_print_format (wv) : & F_8_0; + const struct fmt_spec *wfmt = cmd->wv ? var_get_print_format (cmd->wv) : & F_8_0; int n_rows = 2;