The RATS Software Forum

dear
when i do the wald test using "test" "restrict" or "exclude"
why can not i get the wald Chi-Squared stat, but F stat
how can i get the wald Chi-Squared stat?

See Section 6.2 of the User's Guide. In particular, see "Applicability of the Testing Instructions" (page 222 of the Version 7 edition).

Regards,
Tom Maycock

moderator wrote:See Section 6.2 of the User's Guide. In particular, see "Applicability of the Testing Instructions" (page 222 of the Version 7 edition).

Regards,
Tom Maycock

thank you Tom

just use "robust" option ?

this Chi-Squared stat is Heteroscedasticity adjustment
how can i get the Chi-Squared stat which is not Heteroscedasticity adjustment?

iloverats wrote:

moderator wrote:See Section 6.2 of the User's Guide. In particular, see "Applicability of the Testing Instructions" (page 222 of the Version 7 edition).

Regards,
Tom Maycock

thank you Tom

just use "robust" option ?

The ROBUST option will give you a completely different covariance matrix. If what you want is a conventional Wald test in chi-squared form for a linear regression, you would need to multiply by

# of restrictions x degrees of freedom / observations

The last two factors corrects for the F-statistic using the degrees of freedom adjusted estimate for the variance (rather than the maximum likelihood one) and first corrects for the F using the difference scaled by the number of restrictions. However, I would generally not recommend using the Wald form when an F is sensible (even when the F doesn't have a finite sample justification). The F form is more conservative and more likely to give something closer to the proper behavior.