2
$\begingroup$

Using glm, I have fitted a model with a gamma response distribution and tested all the model diagnostics so everything looks to be a good fit. The only problem is that the model's deviances point to the inverse link function being the most appropriate, however, I not too sure how to interpret the parameters. If I was using a log link, I would use the exponential function on the estimates of beta to get the log odds. For the inverse link function what is the best way to interpret the estimates of beta?

$\endgroup$
  • 1
    $\begingroup$ I'm confused by your question. If you're using a Gamma glm, where do log-odds come in at all? Log odds of what? What is your response? $\endgroup$ – Glen_b Nov 11 '13 at 23:56
  • $\begingroup$ You should tell us more about your modeling context, it is difficult to interpret parameters "in the abstract". Generally, you should not choose link function only based on deviance (fit). You must also think about why you ar molling, what yuou want to ,larn from a model! The inverse link is in general not that easy to interpret, and in many cases, using a log link will b more useful. A log link give a more direct interpretation of parameters! $\endgroup$ – kjetil b halvorsen Jun 30 '14 at 19:15

Your Answer

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

Browse other questions tagged or ask your own question.