projects
/
pspp
/ blobdiff
commit
grep
author
committer
pickaxe
?
search:
re
summary
|
shortlog
|
log
|
commit
|
commitdiff
|
tree
raw
|
inline
| side by side
Fix some typos (found by codespell)
[pspp]
/
src
/
language
/
stats
/
roc.c
diff --git
a/src/language/stats/roc.c
b/src/language/stats/roc.c
index 471b946ee817b17124836b4f601d9358fbd833cf..7d13c2c2cb7cf32c13c9d5c5051cf1328441517a 100644
(file)
--- a/
src/language/stats/roc.c
+++ b/
src/language/stats/roc.c
@@
-49,6
+49,7
@@
struct cmd_roc
const struct variable *state_var;
union value state_value;
const struct variable *state_var;
union value state_value;
+ size_t state_var_width;
/* Plot the roc curve */
bool curve;
/* Plot the roc curve */
bool curve;
@@
-94,6
+95,7
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
roc.neg = roc.neg_weighted = 0;
roc.dict = dataset_dict (ds);
roc.state_var = NULL;
roc.neg = roc.neg_weighted = 0;
roc.dict = dataset_dict (ds);
roc.state_var = NULL;
+ roc.state_var_width = -1;
lex_match (lexer, T_SLASH);
if (!parse_variables_const (lexer, dict, &roc.vars, &roc.n_vars,
lex_match (lexer, T_SLASH);
if (!parse_variables_const (lexer, dict, &roc.vars, &roc.n_vars,
@@
-106,13
+108,18
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
}
roc.state_var = parse_variable (lexer, dict);
}
roc.state_var = parse_variable (lexer, dict);
+ if (! roc.state_var)
+ {
+ goto error;
+ }
if ( !lex_force_match (lexer, T_LPAREN))
{
goto error;
}
if ( !lex_force_match (lexer, T_LPAREN))
{
goto error;
}
- value_init (&roc.state_value, var_get_width (roc.state_var));
+ roc.state_var_width = var_get_width (roc.state_var);
+ value_init (&roc.state_value, roc.state_var_width);
parse_value (lexer, &roc.state_value, roc.state_var);
parse_value (lexer, &roc.state_value, roc.state_var);
@@
-121,7
+128,6
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
goto error;
}
goto error;
}
-
while (lex_token (lexer) != T_ENDCMD)
{
lex_match (lexer, T_SLASH);
while (lex_token (lexer) != T_ENDCMD)
{
lex_match (lexer, T_SLASH);
@@
-154,8
+160,10
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
if (lex_match (lexer, T_LPAREN))
{
roc.reference = true;
if (lex_match (lexer, T_LPAREN))
{
roc.reference = true;
- lex_force_match_id (lexer, "REFERENCE");
- lex_force_match (lexer, T_RPAREN);
+ if (! lex_force_match_id (lexer, "REFERENCE"))
+ goto error;
+ if (! lex_force_match (lexer, T_RPAREN))
+ goto error;
}
}
else if (lex_match_id (lexer, "NONE"))
}
}
else if (lex_match_id (lexer, "NONE"))
@@
-195,7
+203,8
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
{
if (lex_match_id (lexer, "CUTOFF"))
{
{
if (lex_match_id (lexer, "CUTOFF"))
{
- lex_force_match (lexer, T_LPAREN);
+ if (! lex_force_match (lexer, T_LPAREN))
+ goto error;
if (lex_match_id (lexer, "INCLUDE"))
{
roc.exclude = MV_SYSTEM;
if (lex_match_id (lexer, "INCLUDE"))
{
roc.exclude = MV_SYSTEM;
@@
-209,11
+218,13
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
lex_error (lexer, NULL);
goto error;
}
lex_error (lexer, NULL);
goto error;
}
- lex_force_match (lexer, T_RPAREN);
+ if (! lex_force_match (lexer, T_RPAREN))
+ goto error;
}
else if (lex_match_id (lexer, "TESTPOS"))
{
}
else if (lex_match_id (lexer, "TESTPOS"))
{
- lex_force_match (lexer, T_LPAREN);
+ if (! lex_force_match (lexer, T_LPAREN))
+ goto error;
if (lex_match_id (lexer, "LARGE"))
{
roc.invert = false;
if (lex_match_id (lexer, "LARGE"))
{
roc.invert = false;
@@
-227,19
+238,24
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
lex_error (lexer, NULL);
goto error;
}
lex_error (lexer, NULL);
goto error;
}
- lex_force_match (lexer, T_RPAREN);
+ if (! lex_force_match (lexer, T_RPAREN))
+ goto error;
}
else if (lex_match_id (lexer, "CI"))
{
}
else if (lex_match_id (lexer, "CI"))
{
- lex_force_match (lexer, T_LPAREN);
- lex_force_num (lexer);
+ if (!lex_force_match (lexer, T_LPAREN))
+ goto error;
+ if (! lex_force_num (lexer))
+ goto error;
roc.ci = lex_number (lexer);
lex_get (lexer);
roc.ci = lex_number (lexer);
lex_get (lexer);
- lex_force_match (lexer, T_RPAREN);
+ if (!lex_force_match (lexer, T_RPAREN))
+ goto error;
}
else if (lex_match_id (lexer, "DISTRIBUTION"))
{
}
else if (lex_match_id (lexer, "DISTRIBUTION"))
{
- lex_force_match (lexer, T_LPAREN);
+ if (!lex_force_match (lexer, T_LPAREN))
+ goto error;
if (lex_match_id (lexer, "FREE"))
{
roc.bi_neg_exp = false;
if (lex_match_id (lexer, "FREE"))
{
roc.bi_neg_exp = false;
@@
-253,7
+269,8
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
lex_error (lexer, NULL);
goto error;
}
lex_error (lexer, NULL);
goto error;
}
- lex_force_match (lexer, T_RPAREN);
+ if (!lex_force_match (lexer, T_RPAREN))
+ goto error;
}
else
{
}
else
{
@@
-269,16
+286,17
@@
cmd_roc (struct lexer *lexer, struct dataset *ds)
}
}
}
}
- if ( ! run_roc (ds, &roc))
+ if ( ! run_roc (ds, &roc))
goto error;
goto error;
- value_destroy (&roc.state_value, var_get_width (roc.state_var));
+ if ( roc.state_var)
+ value_destroy (&roc.state_value, roc.state_var_width);
free (roc.vars);
return CMD_SUCCESS;
error:
if ( roc.state_var)
free (roc.vars);
return CMD_SUCCESS;
error:
if ( roc.state_var)
- value_destroy (&roc.state_value,
var_get_width (roc.state_var)
);
+ value_destroy (&roc.state_value,
roc.state_var_width
);
free (roc.vars);
return CMD_FAILURE;
}
free (roc.vars);
return CMD_FAILURE;
}
@@
-330,10
+348,10
@@
dump_casereader (struct casereader *reader)
#endif
#endif
-/*
+/*
Return true iff the state variable indicates that C has positive actual state.
Return true iff the state variable indicates that C has positive actual state.
- As a side effect, this function also accumulates the roc->{pos,neg} and
+ As a side effect, this function also accumulates the roc->{pos,neg} and
roc->{pos,neg}_weighted counts.
*/
static bool
roc->{pos,neg}_weighted counts.
*/
static bool
@@
-366,7
+384,7
@@
match_positives (const struct ccase *c, void *aux)
#define N_EQ 1
#define N_PRED 2
#define N_EQ 1
#define N_PRED 2
-/* Some intermediate state for calculating the cutpoints and the
+/* Some intermediate state for calculating the cutpoints and the
standard error values */
struct roc_state
{
standard error values */
struct roc_state
{
@@
-376,7
+394,7
@@
struct roc_state
double n2; /* total weight of negatives */
/* intermediates for standard error */
double n2; /* total weight of negatives */
/* intermediates for standard error */
- double q1hat;
+ double q1hat;
double q2hat;
/* intermediates for cutpoints */
double q2hat;
/* intermediates for cutpoints */
@@
-387,18
+405,18
@@
struct roc_state
double max;
};
double max;
};
-/*
+/*
Return a new casereader based upon CUTPOINT_RDR.
The number of "positive" cases are placed into
the position TRUE_INDEX, and the number of "negative" cases
into FALSE_INDEX.
Return a new casereader based upon CUTPOINT_RDR.
The number of "positive" cases are placed into
the position TRUE_INDEX, and the number of "negative" cases
into FALSE_INDEX.
- POS_COND and RESULT determine the semantics of what is
+ POS_COND and RESULT determine the semantics of what is
"positive".
WEIGHT is the value of a single count.
*/
static struct casereader *
accumulate_counts (struct casereader *input,
"positive".
WEIGHT is the value of a single count.
*/
static struct casereader *
accumulate_counts (struct casereader *input,
- double result, double weight,
+ double result, double weight,
bool (*pos_cond) (double, double),
int true_index, int false_index)
{
bool (*pos_cond) (double, double),
int true_index, int false_index)
{
@@
-451,7
+469,7
@@
static void output_roc (struct roc_state *rs, const struct cmd_roc *roc);
CUTPOINT_RDR accordingly. TRUE_INDEX and FALSE_INDEX are the indices
which receive these values. POS_COND is the condition defining true
and false.
CUTPOINT_RDR accordingly. TRUE_INDEX and FALSE_INDEX are the indices
which receive these values. POS_COND is the condition defining true
and false.
-
+
3. CC is filled with the cumulative weight of all cases of READER.
*/
static struct casereader *
3. CC is filled with the cumulative weight of all cases of READER.
*/
static struct casereader *
@@
-459,7
+477,7
@@
process_group (const struct variable *var, struct casereader *reader,
bool (*pred) (double, double),
const struct dictionary *dict,
double *cc,
bool (*pred) (double, double),
const struct dictionary *dict,
double *cc,
- struct casereader **cutpoint_rdr,
+ struct casereader **cutpoint_rdr,
bool (*pos_cond) (double, double),
int true_index,
int false_index)
bool (*pos_cond) (double, double),
int true_index,
int false_index)
@@
-471,7
+489,7
@@
process_group (const struct variable *var, struct casereader *reader,
const int weight_idx = w ? var_get_case_index (w) :
caseproto_get_n_widths (casereader_get_proto (r1)) - 1;
const int weight_idx = w ? var_get_case_index (w) :
caseproto_get_n_widths (casereader_get_proto (r1)) - 1;
-
+
struct ccase *c1;
struct casereader *rclone = casereader_clone (r1);
struct ccase *c1;
struct casereader *rclone = casereader_clone (r1);
@@
-482,7
+500,7
@@
process_group (const struct variable *var, struct casereader *reader,
proto = caseproto_add_width (proto, 0);
proto = caseproto_add_width (proto, 0);
proto = caseproto_add_width (proto, 0);
proto = caseproto_add_width (proto, 0);
- wtr = autopaging_writer_create (proto);
+ wtr = autopaging_writer_create (proto);
*cc = 0;
*cc = 0;
@@
-528,7
+546,7
@@
process_group (const struct variable *var, struct casereader *reader,
casereader_destroy (r2);
}
casereader_destroy (r2);
}
-
+
casereader_destroy (r1);
casereader_destroy (rclone);
casereader_destroy (r1);
casereader_destroy (rclone);
@@
-539,7
+557,7
@@
process_group (const struct variable *var, struct casereader *reader,
/* Some more indeces into case data */
#define N_POS_EQ 1 /* number of positive cases with values equal to n */
/* Some more indeces into case data */
#define N_POS_EQ 1 /* number of positive cases with values equal to n */
-#define N_POS_GT 2 /* number of postive cases with values greater than n */
+#define N_POS_GT 2 /* number of pos
i
tive cases with values greater than n */
#define N_NEG_EQ 3 /* number of negative cases with values equal to n */
#define N_NEG_LT 4 /* number of negative cases with values less than n */
#define N_NEG_EQ 3 /* number of negative cases with values equal to n */
#define N_NEG_LT 4 /* number of negative cases with values less than n */
@@
-617,7
+635,7
@@
append_cutpoint (struct casewriter *writer, double cutpoint)
}
}
-/*
+/*
Create and initialise the rs[x].cutpoint_rdr casereaders. That is, the readers will
be created with width 5, ready to take the values (cutpoint, ROC_TP, ROC_FN, ROC_TN, ROC_FP), and the
reader will be populated with its final number of cases.
Create and initialise the rs[x].cutpoint_rdr casereaders. That is, the readers will
be created with width 5, ready to take the values (cutpoint, ROC_TP, ROC_FN, ROC_TN, ROC_FP), and the
reader will be populated with its final number of cases.
@@
-658,7
+676,7
@@
prepare_cutpoints (struct cmd_roc *roc, struct roc_state *rs, struct casereader
{
for (i = 0 ; i < roc->n_vars; ++i)
{
{
for (i = 0 ; i < roc->n_vars; ++i)
{
- const union value *v = case_data (c, roc->vars[i]);
+ const union value *v = case_data (c, roc->vars[i]);
const double result = v->f;
if ( mv_is_value_missing (var_get_missing_values (roc->vars[i]), v, roc->exclude))
const double result = v->f;
if ( mv_is_value_missing (var_get_missing_values (roc->vars[i]), v, roc->exclude))
@@
-724,7
+742,7
@@
do_roc (struct cmd_roc *roc, struct casereader *reader, struct dictionary *dict)
/* Separate the positive actual state cases from the negative ones */
/* Separate the positive actual state cases from the negative ones */
- positives =
+ positives =
casereader_create_filter_func (input,
match_positives,
NULL,
casereader_create_filter_func (input,
match_positives,
NULL,
@@
-732,7
+750,7
@@
do_roc (struct cmd_roc *roc, struct casereader *reader, struct dictionary *dict)
neg_wtr);
n_proto = caseproto_create ();
neg_wtr);
n_proto = caseproto_create ();
-
+
n_proto = caseproto_add_width (n_proto, 0);
n_proto = caseproto_add_width (n_proto, 0);
n_proto = caseproto_add_width (n_proto, 0);
n_proto = caseproto_add_width (n_proto, 0);
n_proto = caseproto_add_width (n_proto, 0);
n_proto = caseproto_add_width (n_proto, 0);
@@
-831,7
+849,7
@@
do_roc (struct cmd_roc *roc, struct casereader *reader, struct dictionary *dict)
n_pos_gt = prev_pos_gt;
case_data_rw_idx (nc, N_POS_GT)->f = n_pos_gt;
}
n_pos_gt = prev_pos_gt;
case_data_rw_idx (nc, N_POS_GT)->f = n_pos_gt;
}
-
+
casewriter_write (w, nc);
prev_pos_gt = n_pos_gt;
}
casewriter_write (w, nc);
prev_pos_gt = n_pos_gt;
}
@@
-856,7
+874,7
@@
do_roc (struct cmd_roc *roc, struct casereader *reader, struct dictionary *dict)
n_neg_lt = prev_neg_lt;
case_data_rw_idx (nc, N_NEG_LT)->f = n_neg_lt;
}
n_neg_lt = prev_neg_lt;
case_data_rw_idx (nc, N_NEG_LT)->f = n_neg_lt;
}
-
+
casewriter_write (w, nc);
prev_neg_lt = n_neg_lt;
}
casewriter_write (w, nc);
prev_neg_lt = n_neg_lt;
}
@@
-910,8
+928,8
@@
do_roc (struct cmd_roc *roc, struct casereader *reader, struct dictionary *dict)
casereader_destroy (r);
case_unref (prev_case);
casereader_destroy (r);
case_unref (prev_case);
- rs[i].auc /= rs[i].n1 * rs[i].n2;
- if ( roc->invert )
+ rs[i].auc /= rs[i].n1 * rs[i].n2;
+ if ( roc->invert )
rs[i].auc = 1 - rs[i].auc;
if ( roc->bi_neg_exp )
rs[i].auc = 1 - rs[i].auc;
if ( roc->bi_neg_exp )
@@
-935,7
+953,7
@@
do_roc (struct cmd_roc *roc, struct casereader *reader, struct dictionary *dict)
subcase_destroy (&down_ordering);
output_roc (rs, roc);
subcase_destroy (&down_ordering);
output_roc (rs, roc);
-
+
for (i = 0 ; i < roc->n_vars; ++i)
casereader_destroy (rs[i].cutpoint_rdr);
for (i = 0 ; i < roc->n_vars; ++i)
casereader_destroy (rs[i].cutpoint_rdr);
@@
-996,7
+1014,7
@@
show_auc (struct roc_state *rs, const struct cmd_roc *roc)
{
tab_text (tbl, 0, 2 + i, TAT_TITLE, var_to_string (roc->vars[i]));
{
tab_text (tbl, 0, 2 + i, TAT_TITLE, var_to_string (roc->vars[i]));
- tab_double (tbl, n_cols - n_fields, 2 + i, 0, rs[i].auc, NULL);
+ tab_double (tbl, n_cols - n_fields, 2 + i, 0, rs[i].auc, NULL
, RC_OTHER
);
if ( roc->print_se )
{
if ( roc->print_se )
{
@@
-1015,22
+1033,22
@@
show_auc (struct roc_state *rs, const struct cmd_roc *roc)
tab_double (tbl, n_cols - 4, 2 + i, 0,
se,
tab_double (tbl, n_cols - 4, 2 + i, 0,
se,
- NULL);
+ NULL
, RC_OTHER
);
ci = 1 - roc->ci / 100.0;
yy = gsl_cdf_gaussian_Qinv (ci, se) ;
tab_double (tbl, n_cols - 2, 2 + i, 0,
rs[i].auc - yy,
ci = 1 - roc->ci / 100.0;
yy = gsl_cdf_gaussian_Qinv (ci, se) ;
tab_double (tbl, n_cols - 2, 2 + i, 0,
rs[i].auc - yy,
- NULL);
+ NULL
, RC_OTHER
);
tab_double (tbl, n_cols - 1, 2 + i, 0,
rs[i].auc + yy,
tab_double (tbl, n_cols - 1, 2 + i, 0,
rs[i].auc + yy,
- NULL);
+ NULL
, RC_OTHER
);
tab_double (tbl, n_cols - 3, 2 + i, 0,
2.0 * gsl_cdf_ugaussian_Q (fabs ((rs[i].auc - 0.5 ) / sd_0_5)),
tab_double (tbl, n_cols - 3, 2 + i, 0,
2.0 * gsl_cdf_ugaussian_Q (fabs ((rs[i].auc - 0.5 ) / sd_0_5)),
- NULL);
+ NULL
, RC_PVALUE
);
}
}
}
}
@@
-1077,11
+1095,11
@@
show_summary (const struct cmd_roc *roc)
tab_text (tbl, 0, 3, TAB_LEFT, _("Negative"));
tab_text (tbl, 0, 3, TAB_LEFT, _("Negative"));
- tab_double (tbl, 1, 2, 0, roc->pos,
&F_8_0
);
- tab_double (tbl, 1, 3, 0, roc->neg,
&F_8_0
);
+ tab_double (tbl, 1, 2, 0, roc->pos,
NULL, RC_INTEGER
);
+ tab_double (tbl, 1, 3, 0, roc->neg,
NULL, RC_INTEGER
);
- tab_double (tbl, 2, 2, 0, roc->pos_weighted,
0
);
- tab_double (tbl, 2, 3, 0, roc->neg_weighted,
0
);
+ tab_double (tbl, 2, 2, 0, roc->pos_weighted,
NULL, RC_OTHER
);
+ tab_double (tbl, 2, 3, 0, roc->neg_weighted,
NULL, RC_OTHER
);
tab_submit (tbl);
}
tab_submit (tbl);
}
@@
-1158,10
+1176,10
@@
show_coords (struct roc_state *rs, const struct cmd_roc *roc)
);
tab_double (tbl, n_cols - 3, x, 0, case_data_idx (cc, ROC_CUTPOINT)->f,
);
tab_double (tbl, n_cols - 3, x, 0, case_data_idx (cc, ROC_CUTPOINT)->f,
- var_get_print_format (roc->vars[i]));
+ var_get_print_format (roc->vars[i])
, RC_OTHER
);
- tab_double (tbl, n_cols - 2, x, 0, se, NULL);
- tab_double (tbl, n_cols - 1, x, 0, 1 - sp, NULL);
+ tab_double (tbl, n_cols - 2, x, 0, se, NULL
, RC_OTHER
);
+ tab_double (tbl, n_cols - 1, x, 0, 1 - sp, NULL
, RC_OTHER
);
}
casereader_destroy (r);
}
casereader_destroy (r);