Log or not

[Yuri Groeneveld]

Dear Michel,

Should I use lognested or nested?

Lognested will give me a nest parameter of lower than 1. For both a mode>destination model as for destination>mode model.

If I used nested it will give me this error: runtimeWarning: overflow encountered in double_scalars rho = num / denom.

Any advise?

Thanks in advance.

Yuri

[Bierlaire Michel]

The contribution to the loglikelihood is the log of the probability, given by lognested. 

“nested” (simply defined as exp(lognested)) is used for simulation, not estimation. 

--
You received this message because you are subscribed to the Google Groups "Biogeme" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biogeme+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/f647a044-fd5a-4ce8-b0e6-e425aecdb76bn%40googlegroups.com.