Hello everyone,

I am not sure if the following is a bug or if it
can be avoided, basically I have a gretl script that adds variables (many) to a
model. The script then tests the AIC value to determine if the
var improved the model. However, I received the following
message on a certain variable be omitted from the model:

################

Null hypothesis: the regression parameters
are zero for the variables

ld_ci_10

ld_ci_15

ld_ci_15

Asymptotic test
statistic:

Chi-square(2) = 4.69312, with p-value = 0.0956979

Chi-square(2) = 4.69312, with p-value = 0.0956979

#################

Then gretcli had a segmentation fault. The
core dump regularly happens with the script on the same data. However, it
sometimes does not seg fault with the above message with another model with
different data. That is, gretlcli will continue without a seg fault even
though the Null Hypothesis messages is produced.

Here is a snip of the script:

####

genr maxVar = max(ld_ci_15);

genr minVar = min(ld_ci_15);

if maxVar != minVar;

add ld_ci_15;

genr aicTmp = $aic;

if aicTmp >= aiken - 1;

omit ld_ci_15;

else;

genr aiken = aicTmp;

endif;

endif;

genr minVar = min(ld_ci_15);

if maxVar != minVar;

add ld_ci_15;

genr aicTmp = $aic;

if aicTmp >= aiken - 1;

omit ld_ci_15;

else;

genr aiken = aicTmp;

endif;

endif;

###

One thing I did notice from the log file is that
all other variables actually showed up in the log on "add" and "omit" however
the var in question (ld_ci_15) did not(see log below):

#############

? genr maxVar = max(ld_ci_15)

Replaced scalar maxVar (ID 1109) = 1

? genr minVar = min(ld_ci_15)

Replaced scalar minVar (ID 1110) = -1

Replaced scalar maxVar (ID 1109) = 1

? genr minVar = min(ld_ci_15)

Replaced scalar minVar (ID 1110) = -1

? add ;

....

Comparison of Model 1917 and Model 1918:

Of the 2 model selection statistics, 0 have improved.

Of the 2 model selection statistics, 0 have improved.

? genr aicTmp = $aic

Replaced scalar aicTmp (ID 1111) = 991.01

Replaced scalar aicTmp (ID 1111) = 991.01

? omit ;

.....

.....

##############

Notice how the variable is not listed in the above
snip. The other previous variables do show up with the add and
omit.

My question is, is there a way to do a test on a
variable to determine if it would produce the message "Null hypothesis: the
regression parameters are zero for the variables" prior to adding/omiting to the
model? Or is the core dump a bug that can be fixed?

Thanks,

John

John