0

There is something that i couldn't actually understand about GLMs. In wikipedia it says "(GLM) is a flexible generalization of ordinary linear regression that allows for response variables that have error distribution models other than a normal distribution".

Does this mean logistic regression has a normal distribution as its error distribution ? I think it should have a binomial one. I couldn't even completely understand what an error distribution is...

The thing is, what i understand with the GLM is that it helps to deal with data that has non-linear relations by using linear models. But the function uses residuals so the distribution shouldn't change much about its responses, so i think the link function is what makes the difference.

Can somebody suggest me some sources for learning these models or clerify the misconception on my head ?

anıl ateşsaçan
  • 1
  • 1
    Hi: I remember this thread being a real eye opener for me when I went over it carefully at the time. Of course, now I'd have to go over it carefully again !!!!! https://stats.stackexchange.com/questions/374452/family-of-glm-represents-the-distribution-of-the-response-variable-or-residuals/374498#374498 – mlofton May 05 '21 at 22:54
  • @mlofton thank you so muchh, it really helped me and made it clear. – anıl ateşsaçan May 06 '21 at 12:12
  • I'm glad it helped. I remember it helping me a lot also. – mlofton May 07 '21 at 13:47

0 Answers0