projects
/
pspp
/ blobdiff
commit
grep
author
committer
pickaxe
?
search:
re
summary
|
shortlog
|
log
|
commit
|
commitdiff
|
tree
raw
|
inline
| side by side
Add code to read character encoding to dissect-sysfile.
[pspp]
/
src
/
language
/
stats
/
frequencies.q
diff --git
a/src/language/stats/frequencies.q
b/src/language/stats/frequencies.q
index 0934fc4afadc85e69c29bc93fbf2354591ccbdea..cd370be821c8db73f2d88da8350613a7ea8c07b0 100644
(file)
--- a/
src/language/stats/frequencies.q
+++ b/
src/language/stats/frequencies.q
@@
-1,5
+1,5
@@
/* PSPP - a program for statistical analysis.
/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2007 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2007
, 2009
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
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
@@
-38,12
+38,10
@@
#include <language/command.h>
#include <language/dictionary/split-file.h>
#include <language/lexer/lexer.h>
#include <language/command.h>
#include <language/dictionary/split-file.h>
#include <language/lexer/lexer.h>
-#include <libpspp/alloc.h>
#include <libpspp/array.h>
#include <libpspp/bit-vector.h>
#include <libpspp/compiler.h>
#include <libpspp/hash.h>
#include <libpspp/array.h>
#include <libpspp/bit-vector.h>
#include <libpspp/compiler.h>
#include <libpspp/hash.h>
-#include <libpspp/magic.h>
#include <libpspp/message.h>
#include <libpspp/misc.h>
#include <libpspp/pool.h>
#include <libpspp/message.h>
#include <libpspp/misc.h>
#include <libpspp/pool.h>
@@
-60,6
+58,7
@@
#include "freq.h"
#include "minmax.h"
#include "freq.h"
#include "minmax.h"
+#include "xalloc.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
#include "gettext.h"
#define _(msgid) gettext (msgid)
@@
-156,8
+155,6
@@
static void add_percentile (double x) ;
static struct percentile *percentiles;
static int n_percentiles;
static struct percentile *percentiles;
static int n_percentiles;
-static int implicit_50th ;
-
/* Groups of statistics. */
#define BI BIT_INDEX
#define frq_default \
/* Groups of statistics. */
#define BI BIT_INDEX
#define frq_default \
@@
-271,7
+268,7
@@
static hsh_compare_func compare_freq_numeric_d, compare_freq_alpha_d;
static void do_piechart(const struct variable *var,
const struct freq_tab *frq_tab);
static void do_piechart(const struct variable *var,
const struct freq_tab *frq_tab);
-
gsl_
histogram *
+
struct
histogram *
freq_tab_to_hist(const struct freq_tab *ft, const struct variable *var);
freq_tab_to_hist(const struct freq_tab *ft, const struct variable *var);
@@
-313,7
+310,7
@@
internal_cmd_frequencies (struct lexer *lexer, struct dataset *ds)
if (!parse_frequencies (lexer, ds, &cmd, NULL))
return CMD_FAILURE;
if (!parse_frequencies (lexer, ds, &cmd, NULL))
return CMD_FAILURE;
- if (cmd.onepage_limit ==
NOT_LONG
)
+ if (cmd.onepage_limit ==
LONG_MIN
)
cmd.onepage_limit = 50;
/* Figure out statistics to calculate. */
cmd.onepage_limit = 50;
/* Figure out statistics to calculate. */
@@
-323,14
+320,14
@@
internal_cmd_frequencies (struct lexer *lexer, struct dataset *ds)
if (cmd.a_statistics[FRQ_ST_ALL])
stats |= frq_all;
if (cmd.sort != FRQ_AVALUE && cmd.sort != FRQ_DVALUE)
if (cmd.a_statistics[FRQ_ST_ALL])
stats |= frq_all;
if (cmd.sort != FRQ_AVALUE && cmd.sort != FRQ_DVALUE)
- stats &= ~
frq_median
;
+ stats &= ~
BIT_INDEX (frq_median)
;
for (i = 0; i < frq_n_stats; i++)
if (cmd.a_statistics[st_name[i].st_indx])
stats |= BIT_INDEX (i);
if (stats & frq_kurt)
for (i = 0; i < frq_n_stats; i++)
if (cmd.a_statistics[st_name[i].st_indx])
stats |= BIT_INDEX (i);
if (stats & frq_kurt)
- stats |=
frq_sekurt
;
+ stats |=
BIT_INDEX (frq_sekurt)
;
if (stats & frq_skew)
if (stats & frq_skew)
- stats |=
frq_seskew
;
+ stats |=
BIT_INDEX (frq_seskew)
;
/* Calculate n_stats. */
n_stats = 0;
/* Calculate n_stats. */
n_stats = 0;
@@
-363,7
+360,14
@@
internal_cmd_frequencies (struct lexer *lexer, struct dataset *ds)
add_percentile (j / (double) cmd.n_ntiles[i]);
}
}
add_percentile (j / (double) cmd.n_ntiles[i]);
}
}
-
+ if (stats & BIT_INDEX (frq_median))
+ {
+ /* Treat the median as the 50% percentile.
+ We output it in the percentiles table as "50 (Median)." */
+ add_percentile (0.5);
+ stats &= ~BIT_INDEX (frq_median);
+ n_stats--;
+ }
/* Do it! */
input = casereader_create_filter_weight (proc_open (ds), dataset_dict (ds),
/* Do it! */
input = casereader_create_filter_weight (proc_open (ds), dataset_dict (ds),
@@
-372,11
+376,11
@@
internal_cmd_frequencies (struct lexer *lexer, struct dataset *ds)
for (; casegrouper_get_next_group (grouper, &group);
casereader_destroy (group))
{
for (; casegrouper_get_next_group (grouper, &group);
casereader_destroy (group))
{
- struct ccase c;
+ struct ccase
*
c;
precalc (group, ds);
precalc (group, ds);
- for (;
casereader_read (group, &c); case_destroy (&
c))
- calc (
&
c, ds);
+ for (;
(c = casereader_read (group)) != NULL; case_unref (
c))
+ calc (c, ds);
postcalc ();
}
ok = casegrouper_destroy (grouper);
postcalc ();
}
ok = casegrouper_destroy (grouper);
@@
-531,13
+535,15
@@
calc (const struct ccase *c, const struct dataset *ds)
static void
precalc (struct casereader *input, struct dataset *ds)
{
static void
precalc (struct casereader *input, struct dataset *ds)
{
- struct ccase c;
+ struct ccase
*
c;
size_t i;
size_t i;
- if (!casereader_peek (input, 0, &c))
- return;
- output_split_file_values (ds, &c);
- case_destroy (&c);
+ c = casereader_peek (input, 0);
+ if (c != NULL)
+ {
+ output_split_file_values (ds, c);
+ case_unref (c);
+ }
pool_destroy (data_pool);
data_pool = pool_create ();
pool_destroy (data_pool);
data_pool = pool_create ();
@@
-601,43
+607,38
@@
postcalc (void)
if ( chart == GFT_HIST)
{
double d[frq_n_stats];
if ( chart == GFT_HIST)
{
double d[frq_n_stats];
- struct normal_curve norm;
- gsl_histogram *hist ;
-
-
- norm.N = vf->tab.valid_cases;
+ struct histogram *hist ;
calc_stats (v, d);
calc_stats (v, d);
- norm.mean = d[frq_mean];
- norm.stddev = d[frq_stddev];
- hist = freq_tab_to_hist(ft,v);
+ hist = freq_tab_to_hist
(ft,v);
- histogram_plot(hist, var_to_string(v), &norm, normal);
+ histogram_plot_n (hist, var_to_string(v),
+ vf->tab.valid_cases,
+ d[frq_mean],
+ d[frq_stddev],
+ normal);
-
gsl_histogram_free(
hist);
+
statistic_destroy ((struct statistic *)
hist);
}
}
-
if ( chart == GFT_PIE)
{
do_piechart(v_variables[i], ft);
}
if ( chart == GFT_PIE)
{
do_piechart(v_variables[i], ft);
}
-
-
cleanup_freq_tab (v);
}
}
/* Returns the comparison function that should be used for
cleanup_freq_tab (v);
}
}
/* Returns the comparison function that should be used for
- sorting a frequency table by FRQ_SORT using VA
R
_TYPE
- va
riabl
es. */
+ sorting a frequency table by FRQ_SORT using VA
L
_TYPE
+ va
lu
es. */
static hsh_compare_func *
static hsh_compare_func *
-get_freq_comparator (int frq_sort, enum va
r_type var
_type)
+get_freq_comparator (int frq_sort, enum va
l_type val
_type)
{
{
- bool is_numeric = va
r_type == VAR
_NUMERIC;
+ bool is_numeric = va
l_type == VAL
_NUMERIC;
switch (frq_sort)
{
case FRQ_AVALUE:
switch (frq_sort)
{
case FRQ_AVALUE:
@@
-761,7
+762,7
@@
frq_custom_variables (struct lexer *lexer, struct dataset *ds, struct cmd_freque
vf->groups = NULL;
vf->width = var_get_width (v);
vf->print = *var_get_print_format (v);
vf->groups = NULL;
vf->width = var_get_width (v);
vf->print = *var_get_print_format (v);
- if (vf->width > MAX_SHORT_STRING && get_algorithm () == COMPATIBLE)
+ if (vf->width > MAX_SHORT_STRING &&
settings_
get_algorithm () == COMPATIBLE)
{
enum fmt_type type = var_get_print_format (v)->type;
vf->width = MAX_SHORT_STRING;
{
enum fmt_type type = var_get_print_format (v)->type;
vf->width = MAX_SHORT_STRING;
@@
-876,11
+877,7
@@
add_percentile (double x)
{
percentiles = pool_nrealloc (syntax_pool, percentiles,
n_percentiles + 1, sizeof *percentiles);
{
percentiles = pool_nrealloc (syntax_pool, percentiles,
n_percentiles + 1, sizeof *percentiles);
-
- if (i < n_percentiles)
- memmove (&percentiles[i + 1], &percentiles[i],
- (n_percentiles - i) * sizeof (struct percentile) );
-
+ insert_element (percentiles, n_percentiles, sizeof *percentiles, i);
percentiles[i].p = x;
n_percentiles++;
}
percentiles[i].p = x;
n_percentiles++;
}
@@
-1236,28
+1233,9
@@
calc_stats (const struct variable *v, double d[frq_n_stats])
double rank;
int i = 0;
int idx;
double rank;
int i = 0;
int idx;
- double *median_value;
/* Calculate percentiles. */
/* Calculate percentiles. */
- /* If the 50th percentile was not explicitly requested then we must
- calculate it anyway --- it's the median */
- median_value = 0 ;
- for (i = 0; i < n_percentiles; i++)
- {
- if (percentiles[i].p == 0.5)
- {
- median_value = &percentiles[i].value;
- break;
- }
- }
-
- if ( 0 == median_value )
- {
- add_percentile (0.5);
- implicit_50th = 1;
- }
-
for (i = 0; i < n_percentiles; i++)
{
percentiles[i].flag = 0;
for (i = 0; i < n_percentiles; i++)
{
percentiles[i].flag = 0;
@@
-1275,7
+1253,7
@@
calc_stats (const struct variable *v, double d[frq_n_stats])
double tp;
if ( percentiles[i].flag2 ) continue ;
double tp;
if ( percentiles[i].flag2 ) continue ;
- if (
get_algorithm
() != COMPATIBLE )
+ if (
settings_get_algorithm
() != COMPATIBLE )
tp =
(ft->valid_cases - 1) * percentiles[i].p;
else
tp =
(ft->valid_cases - 1) * percentiles[i].p;
else
@@
-1326,7
+1304,7
@@
calc_stats (const struct variable *v, double d[frq_n_stats])
double s;
double dummy;
double s;
double dummy;
- if (
get_algorithm
() != COMPATIBLE )
+ if (
settings_get_algorithm
() != COMPATIBLE )
{
s = modf((ft->valid_cases - 1) * percentiles[i].p , &dummy);
}
{
s = modf((ft->valid_cases - 1) * percentiles[i].p , &dummy);
}
@@
-1337,9
+1315,6
@@
calc_stats (const struct variable *v, double d[frq_n_stats])
percentiles[i].value = percentiles[i].x1 +
( percentiles[i].x2 - percentiles[i].x1) * s ;
percentiles[i].value = percentiles[i].x1 +
( percentiles[i].x2 - percentiles[i].x1) * s ;
-
- if ( percentiles[i].p == 0.50)
- median_value = &percentiles[i].value;
}
}
@@
-1376,7
+1351,6
@@
calc_stats (const struct variable *v, double d[frq_n_stats])
d[frq_max] = ft->valid[ft->n_valid - 1].value[0].f;
d[frq_mode] = X_mode;
d[frq_range] = d[frq_max] - d[frq_min];
d[frq_max] = ft->valid[ft->n_valid - 1].value[0].f;
d[frq_mode] = X_mode;
d[frq_range] = d[frq_max] - d[frq_min];
- d[frq_median] = *median_value;
d[frq_sum] = d[frq_mean] * W;
d[frq_stddev] = sqrt (d[frq_variance]);
d[frq_semean] = d[frq_stddev] / sqrt (W);
d[frq_sum] = d[frq_mean] * W;
d[frq_stddev] = sqrt (d[frq_variance]);
d[frq_semean] = d[frq_stddev] / sqrt (W);
@@
-1393,11
+1367,6
@@
dump_statistics (const struct variable *v, int show_varname)
struct tab_table *t;
int i, r;
struct tab_table *t;
int i, r;
- int n_explicit_percentiles = n_percentiles;
-
- if ( implicit_50th && n_percentiles > 0 )
- --n_percentiles;
-
if (var_is_alpha (v))
return;
ft = &get_var_freqs (v)->tab;
if (var_is_alpha (v))
return;
ft = &get_var_freqs (v)->tab;
@@
-1409,7
+1378,7
@@
dump_statistics (const struct variable *v, int show_varname)
}
calc_stats (v, stat_value);
}
calc_stats (v, stat_value);
- t = tab_create (3, n_stats + n_
explicit_
percentiles + 2, 0);
+ t = tab_create (3, n_stats + n_percentiles + 2, 0);
tab_dim (t, tab_natural_dimensions);
tab_box (t, TAL_1, TAL_1, -1, -1 , 0 , 0 , 2, tab_nr(t) - 1) ;
tab_dim (t, tab_natural_dimensions);
tab_box (t, TAL_1, TAL_1, -1, -1 , 0 , 0 , 2, tab_nr(t) - 1) ;
@@
-1437,14
+1406,17
@@
dump_statistics (const struct variable *v, int show_varname)
tab_float(t, 2, 1, TAB_NONE, ft->total_cases - ft->valid_cases, 11, 0);
tab_float(t, 2, 1, TAB_NONE, ft->total_cases - ft->valid_cases, 11, 0);
- for (i = 0; i < n_
explicit_
percentiles; i++, r++)
+ for (i = 0; i < n_percentiles; i++, r++)
{
if ( i == 0 )
{
tab_text (t, 0, r, TAB_LEFT | TAT_TITLE, _("Percentiles"));
}
{
if ( i == 0 )
{
tab_text (t, 0, r, TAB_LEFT | TAT_TITLE, _("Percentiles"));
}
- tab_float (t, 1, r, TAB_LEFT, percentiles[i].p * 100, 3, 0 );
+ if (percentiles[i].p == 0.5)
+ tab_text (t, 1, r, TAB_LEFT, _("50 (Median)"));
+ else
+ tab_float (t, 1, r, TAB_LEFT, percentiles[i].p * 100, 3, 0);
tab_float (t, 2, r, TAB_NONE, percentiles[i].value, 11, 3);
}
tab_float (t, 2, r, TAB_NONE, percentiles[i].value, 11, 3);
}
@@
-1461,14
+1433,14
@@
dump_statistics (const struct variable *v, int show_varname)
/* Create a gsl_histogram from a freq_tab */
/* Create a gsl_histogram from a freq_tab */
-
gsl_
histogram *
-freq_tab_to_hist(const struct freq_tab *ft, const struct variable *var)
+
struct
histogram *
+freq_tab_to_hist
(const struct freq_tab *ft, const struct variable *var)
{
int i;
double x_min = DBL_MAX;
double x_max = -DBL_MAX;
{
int i;
double x_min = DBL_MAX;
double x_max = -DBL_MAX;
-
gsl_histogram
*hist;
+
struct statistic
*hist;
const double bins = 11;
struct hsh_iterator hi;
const double bins = 11;
struct hsh_iterator hi;
@@
-1485,15
+1457,15
@@
freq_tab_to_hist(const struct freq_tab *ft, const struct variable *var)
if ( frq->value[0].f > x_max ) x_max = frq->value[0].f ;
}
if ( frq->value[0].f > x_max ) x_max = frq->value[0].f ;
}
- hist = histogram_create(bins, x_min, x_max);
+ hist = histogram_create
(bins, x_min, x_max);
for( i = 0 ; i < ft->n_valid ; ++i )
{
frq = &ft->valid[i];
for( i = 0 ; i < ft->n_valid ; ++i )
{
frq = &ft->valid[i];
-
gsl_histogram_accumulate(
hist, frq->value[0].f, frq->count);
+
histogram_add ((struct histogram *)
hist, frq->value[0].f, frq->count);
}
}
- return hist;
+ return
(struct histogram *)
hist;
}
}
@@
-1523,7
+1495,8
@@
freq_tab_to_slice_array(const struct freq_tab *frq_tab,
{
const struct freq *frq = &frq_tab->valid[i];
{
const struct freq *frq = &frq_tab->valid[i];
- slices[i].label = var_get_value_name (var, frq->value);
+ ds_init_empty (&slices[i].label);
+ var_append_value_name (var, frq->value, &slices[i].label);
slices[i].magnetude = frq->count;
}
slices[i].magnetude = frq->count;
}
@@
-1537,12
+1510,17
@@
static void
do_piechart(const struct variable *var, const struct freq_tab *frq_tab)
{
struct slice *slices;
do_piechart(const struct variable *var, const struct freq_tab *frq_tab)
{
struct slice *slices;
- int n_slices;
+ int n_slices
, i
;
slices = freq_tab_to_slice_array(frq_tab, var, &n_slices);
piechart_plot(var_to_string(var), slices, n_slices);
slices = freq_tab_to_slice_array(frq_tab, var, &n_slices);
piechart_plot(var_to_string(var), slices, n_slices);
+ for (i = 0 ; i < n_slices ; ++i )
+ {
+ ds_destroy (&slices[i].label);
+ }
+
free(slices);
}
free(slices);
}