www.cmm.bristol.ac.uk/forum

Hello,

I'm trying to run a three-level model using a normal distribution with a log link. However, I receive the following error message: "==: operator invalid
r(198);"

I also receive this message when I try to run a three level model using data available on your website in the "Examples" page in order to double-check if the issue is with my data.

I have no problem running the model if I use another distribution e.g. the Poisson with a log link either with my own data or your data.

Is it possible to use a log link with a normal distribution or could something else be causing the problem?

Many thanks,
Valerie

MLwiN currently only allows the identity link for Normal models, so if you hadn't got the error in Stata it still wouldn't have fitted the model. We should probably provide a more informative error message when this is attempted, so I will try to update this in later versions.

OK - good to know. Thanks for your prompt response.

Valerie

Hi Valerie,

Can you not just log transform your dependent variable and then proceed as normal

. generate newdepvar = log(olddepvar)

. runmlwin newdepvar...

Best wishes

George