[R] Trust in a glm.nb model results with an itereation limit reached

[David Winsemius]

On Feb 10, 2012, at 7:14 AM, Lucas wrote:

> Hello to everyone.
> I'm fitting a glm.nb model to a count data. I'm using about 8  
> predictive
> variables.
> Once a run the script I do get a result but it tells me that the  
> iteration
> limit has been reached.

Most regression functions have a control parameter that can increase  
the maximum number of iterations, but you provided no code so all is  
guesswork from here.

> So, can i trust the results given by the model?

Probably not. Some functions will not even record estimates if they  
have failed the convergence tests.

> Could it be a multicollinearity problem?

Could be.

>
> Thank you for taking the time to help me.

-- 

David Winsemius, MD
West Hartford, CT