2
$\begingroup$

I am using the logistf package available for SPPS to carry out a firth logistic regression, and have results relating to the coefficents, standard errors and p-values associated with each predictor. I would now like to calculate the R square statistic associated with the model (preferably McFadden’s) in order to make some assessment of model fit, and to contrast model fit when particular variables are added in or removed. However, the output from the logistf package doesn’t provide me with this, nor does it seem to provide me with enough information to calculate this statistic manually. How would I acquire this statistic given the information I have available to me? (I have included pictures of this below).

This is the only output I get from the SPSS logistf output

$\endgroup$
1
1
$\begingroup$

In the current version of the STATS_FIRTHLOG extension, there is no way to get at the log-likelihood values you would need to calculate McFadden's or other pseudo-R2 values. After verifying that the R logistf package did make that information available, I contacted the author of the extension, and he's updated his files. Keep an eye on the SPSS Statistics Extension Hub for a new version of STATS_FIRTHLOG. When that's available, update your version and you should have access to the log-likelihood values necessary to calculate these kinds of statistics.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.