X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Foneway.c;h=aa8a255211192852f78318940d68179e046ec28c;hb=c1ecb26e96f0abc014a374081e7c2f985b2ca112;hp=f481c69dc13c765cfd55cbc346ca2089a15df7f2;hpb=af097b7430b4080eec4399df6898a9e97d17d0b4;p=pspp diff --git a/src/language/stats/oneway.c b/src/language/stats/oneway.c index f481c69dc1..aa8a255211 100644 --- a/src/language/stats/oneway.c +++ b/src/language/stats/oneway.c @@ -102,6 +102,15 @@ struct oneway_spec }; +/* Per category data */ +struct descriptive_data +{ + const struct variable *var; + struct moments1 *mom; + + double minimum; + double maximum; +}; /* Workspace variable for each dependent variable */ struct per_var_ws @@ -114,7 +123,6 @@ struct per_var_ws int n_groups; - double cc; double mse; }; @@ -314,15 +322,6 @@ free_double (void *value_, const void *aux UNUSED) static void postcalc (const struct oneway_spec *cmd); static void precalc (const struct oneway_spec *cmd); -struct descriptive_data -{ - const struct variable *var; - struct moments1 *mom; - - double minimum; - double maximum; -}; - static struct descriptive_data * dd_create (const struct variable *var) { @@ -348,7 +347,9 @@ makeit (void *aux1, void *aux2 UNUSED) } static void -updateit (void *user_data, const struct variable *wv, +updateit (void *user_data, + enum mv_class exclude, + const struct variable *wv, const struct variable *catvar UNUSED, const struct ccase *c, void *aux1, void *aux2) @@ -361,9 +362,12 @@ updateit (void *user_data, const struct variable *wv, struct descriptive_data *dd_total = aux2; - double weight = 1.0; - if (wv) - weight = case_data (c, wv)->f; + double weight; + + if ( var_is_value_missing (varp, valx, exclude)) + return; + + weight = wv != NULL ? case_data (c, wv)->f : 1.0; moments1_add (dd->mom, valx->f, weight); if (valx->f * weight < dd->minimum) @@ -400,17 +404,13 @@ run_oneway (const struct oneway_spec *cmd, struct ccase *c; struct oneway_workspace ws; - - { - ws.vws = xmalloc (cmd->n_vars * sizeof (*ws.vws)); - ws.dd_total = xmalloc (sizeof (struct descriptive_data) * cmd->n_vars); + ws.actual_number_of_groups = 0; + ws.vws = xmalloc (cmd->n_vars * sizeof (*ws.vws)); + ws.dd_total = xmalloc (sizeof (struct descriptive_data) * cmd->n_vars); - for (v = 0 ; v < cmd->n_vars; ++v) - { - ws.dd_total[v] = dd_create (cmd->vars[v]); - } - } + for (v = 0 ; v < cmd->n_vars; ++v) + ws.dd_total[v] = dd_create (cmd->vars[v]); for (v = 0; v < cmd->n_vars; ++v) { @@ -423,7 +423,6 @@ run_oneway (const struct oneway_spec *cmd, ws.vws[v].cov = covariance_2pass_create (1, &cmd->vars[v], cats, cmd->wv, cmd->exclude); - ws.vws[v].cc = 0; } c = casereader_peek (input, 0); @@ -470,22 +469,21 @@ 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); - } - + struct per_var_ws *pvw = &ws.vws[i]; const struct variable *v = cmd->vars[i]; - const union value *val = case_data (c, v); - struct group_proc *gp = group_proc_get (cmd->vars[i]); struct hsh_table *group_hash = gp->group_hash; - struct group_statistics *gs; + if ( MISS_ANALYSIS == cmd->missing_type) + { + if ( var_is_value_missing (v, val, cmd->exclude)) + continue; + } + + covariance_accumulate_pass1 (pvw->cov, c); + gs = hsh_find (group_hash, indep_val ); if ( ! gs ) @@ -539,6 +537,15 @@ run_oneway (const struct oneway_spec *cmd, for (i = 0; i < cmd->n_vars; ++i) { struct per_var_ws *pvw = &ws.vws[i]; + const struct variable *v = cmd->vars[i]; + const union value *val = case_data (c, v); + + if ( MISS_ANALYSIS == cmd->missing_type) + { + if ( var_is_value_missing (v, val, cmd->exclude)) + continue; + } + covariance_accumulate_pass2 (pvw->cov, c); } } @@ -550,8 +557,13 @@ run_oneway (const struct oneway_spec *cmd, gsl_matrix *cm = covariance_calculate_unnormalized (pvw->cov); const struct categoricals *cats = covariance_get_categoricals (pvw->cov); + double n; + moments1_calculate (ws.dd_total[v]->mom, &n, NULL, NULL, NULL, NULL); + pvw->sst = gsl_matrix_get (cm, 0, 0); + // gsl_matrix_fprintf (stdout, cm, "%g "); + reg_sweep (cm, 0); pvw->sse = gsl_matrix_get (cm, 0, 0); @@ -560,16 +572,20 @@ run_oneway (const struct oneway_spec *cmd, pvw->n_groups = categoricals_total (cats); - pvw->mse = (pvw->sst - pvw->ssa) / (pvw->cc - pvw->n_groups); + pvw->mse = (pvw->sst - pvw->ssa) / (n - pvw->n_groups); } postcalc (cmd); + for (v = 0; v < cmd->n_vars; ++v) { struct categoricals *cats = covariance_get_categoricals (ws.vws[v].cov); categoricals_done (cats); + + if (categoricals_total (cats) > ws.actual_number_of_groups) + ws.actual_number_of_groups = categoricals_total (cats); } if ( cmd->stats & STATS_HOMOGENEITY ) @@ -578,8 +594,6 @@ run_oneway (const struct oneway_spec *cmd, casereader_destroy (input); - ws.actual_number_of_groups = hsh_count (ws.group_hash); - if (!taint_has_tainted_successor (taint)) output_oneway (cmd, &ws); @@ -695,14 +709,12 @@ output_oneway (const struct oneway_spec *cmd, struct oneway_workspace *ws) show_anova_table (cmd, ws); - if (ll_count (&cmd->contrast_list) > 0) { show_contrast_coeffs (cmd, ws); show_contrast_tests (cmd, ws); } - /* Clean up */ for (i = 0; i < cmd->n_vars; ++i ) { @@ -746,12 +758,17 @@ show_anova_table (const struct oneway_spec *cmd, const struct oneway_workspace * for (i = 0; i < cmd->n_vars; ++i) { + double n; + double df1, df2; + double msa; + const char *s = var_to_string (cmd->vars[i]); const struct per_var_ws *pvw = &ws->vws[i]; - const double df1 = pvw->n_groups - 1; - const double df2 = pvw->cc - pvw->n_groups; - const double msa = pvw->ssa / df1; - const char *s = var_to_string (cmd->vars[i]); + moments1_calculate (ws->dd_total[i]->mom, &n, NULL, NULL, NULL, NULL); + + df1 = pvw->n_groups - 1; + df2 = n - pvw->n_groups; + msa = pvw->ssa / df1; 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")); @@ -771,7 +788,7 @@ show_anova_table (const struct oneway_spec *cmd, const struct oneway_workspace * /* 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); + tab_fixed (t, 3, i * 3 + 3, 0, n - 1, 4, 0); /* Mean Squares */ tab_double (t, 4, i * 3 + 1, TAB_RIGHT, msa, NULL); @@ -987,19 +1004,23 @@ show_homogeneity (const struct oneway_spec *cmd, const struct oneway_workspace * for (v = 0; v < cmd->n_vars; ++v) { + double n; + + const struct per_var_ws *pvw = &ws->vws[v]; - const struct categoricals *cats = covariance_get_categoricals (pvw->cov); const struct variable *var = cmd->vars[v]; const struct group_proc *gp = group_proc_get (cmd->vars[v]); const char *s = var_to_string (var); - - const double df1 = pvw->n_groups - 1; - const double df2 = pvw->cc - pvw->n_groups; + double df1, df2; double F = gp->levene; - tab_text (t, 0, v + 1, TAB_LEFT | TAT_TITLE, s); + moments1_calculate (ws->dd_total[v]->mom, &n, NULL, NULL, NULL, NULL); + df1 = pvw->n_groups - 1; + df2 = n - pvw->n_groups; + + tab_text (t, 0, v + 1, TAB_LEFT | TAT_TITLE, s); tab_double (t, 1, v + 1, TAB_RIGHT, F, NULL); tab_fixed (t, 2, v + 1, TAB_RIGHT, df1, 8, 0); @@ -1216,13 +1237,13 @@ show_contrast_tests (const struct oneway_spec *cmd, const struct oneway_workspac { double n, mean, variance; const struct descriptive_data *dd = categoricals_get_user_data_by_subscript (cats, ci); - - moments1_calculate (dd->mom, &n, &mean, &variance, NULL, NULL); - struct coeff_node *cn = ll_data (coeffi, struct coeff_node, ll); const double coef = cn->coeff; + double winv ; + + moments1_calculate (dd->mom, &n, &mean, &variance, NULL, NULL); - const double winv = variance / n; + winv = variance / n; contrast_value += coef * mean;