|
|
Dear all,
I ran a multinomial logit model with SPSS 14 using nomreg. In the
output I found
McFadden's R2 which should be calculated on the basis of
1 - lnL(full model)/lnL(intercept only model).
If I use the numbers of the above LnLs which are printed in SPSS output and
calculate the R2 myself I get a different result than SPSS. But if I run nomreg
in the batch mode (with print=kernel) and I calculate the R2 myself again
I get the same result like SPSS.
If I compute Nagelkerke's R2 it's just the other way around.
So SPSS seems to use the Log-Likelihood-function in order to compute
Nagelkerke's R2 and the kernel of the Log-Likelihood-function in order to
compute the McFadden's R2.
Why does SPSS use different types of Log-Likelihood-functions?
Thank you very much for any hints.
Michael
|
Locked
