Re: GLM: a bernoulli or a binomial response?

On 12/26/2006 9:30 AM, Aleik wrote:
I have a data frame showing a fairly large number of observations,
whose response is "yes" or "no", and there are 2 numerical covariates
and 6 yes-or-no factors supposedly predicting the response. My problem
is that in trying to investigate the predictive power of each variable
alone, there doesn't seem to be a lot of lee-way (I can't easily see
the relationship if there is one).

At first sight the response "yes" or "no" seems to be bernoulli, but
perhaps analysis is more appropriate on a binomial response, or
binomial proportion response, as there are quite a lot of observations?
But there seem to be too many explanatory variables to do it this way.
Can anyone shed some light on which probability distribution I should
use?

I'm really stuck, and I would appreciate help greatly.

Aleik.

You should look into Generalized (not General) Linear Models, which are designed to handle response variable(s) that have one of a large number of distributions ... including your case, where the response is indeed binomial.

--
Paige Miller
pmiller5@xxxxxxxxxxxxxxxx

It's nothing until I call it -- Bill Klem, NL Umpire
If you get the choice to sit it out or dance,
I hope you dance -- Lee Ann Womack
.

Relevant Pages