Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am working on my thesis analysis, and I have some error data that's right-skewed. I log-transformed it and ran glm on it (gaussian, identity in R) weighted by sample size, and my data is still over-dispersed. I keep seeing comments about using the quasi family, but I'm not sure I completely understand what I'm doing. I tried it out, and my question is, can I use quasi with the log-transformed data and still set my variance equal to my mean (or mean squared), or is that like a double variance stabilizer? If I run the data with quasi, link identity, variance = mean in R, I get approximately the same results, except the data appears to by neither under nor over-dispersed.


share|improve this question
Can you provide more details about your situation? What kind of data do you have? What does it mean to have over-dispersion if your response distribution is Gaussian? In general, I would not use the term "generalized linear model" to describe a model w/ an identity link & a Gaussian response (although it isn't wrong), I'd just say 'regression'. – gung Nov 23 '12 at 23:36
Erm... how is a gaussian 'over-dispersed'? Relative to what? – Glen_b Nov 24 '12 at 0:33

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.