[R] reference category in binomal glm

[David Winsemius]

On May 27, 2014, at 1:51 AM, Xebar Saram wrote:

> Hi all
> 
> i know this is probably a silly question but im wondering what is the
> 'reference' category when you run a binomal glm. that is my outcome/DV is
> 0,1 and i run a regression and get coefficients. do the coefficients refer
> to the probability to get 0 or 1?

No, would be my quick answer.

With logistic regression, i.e., "binomial glm", the coefficients are the log-odds of an outcome of 1 minus the log-odds of an outcome of 0 for a one unit change or difference in the predictor whose effect is being estimated. This is known log of the "odds ratio" or the logit (since subtraction on the log-odds scale is equivalent to division on the odds-ratio scale.)

When you ask about a "reference level", it is unclear whether you refer to the reference level for factor coefficients or your are talking about the reference level for the response.

With the use of 'predict.glm' (called form by the use of 'predict(.)" with a glm fit), you have a choice of specifying whether you want the "type" to be "response" to return the probabilities or "terms" to return the usual coefficients on the log-odds scale.

See:

?predict.glm

> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

David Winsemius
Alameda, CA, USA