1

I have a problem where the expected regression output should be "loan amount". The expected output should be between 2 values (EX: 100K and 1000K)

I was thinking of using a Gamma GLM since the outcome variable is positive and skewed.

So my question are:

  1. Is there any other suggestions for different GLMs that meet this requirement?
  2. What are the pros and cons for using Gamma GLM vs linear regression using transformed predictor variables (most of them are financial data so they are log normally distributed) ?
  3. If I were to use Gamma GLM, what type of transformations/checks do I need before fitting the data?
Mohamed Gamal Hamed
  • 21
  • 2
  • Usual linear regression assumes that variance is constant, Gamma models assume that the coefficient of variation is constant. So, if the later is a better fit than the first, that is a good reason. See also https://stats.stackexchange.com/questions/190763/how-to-decide-which-glm-family-to-use, https://stats.stackexchange.com/questions/38282/how-do-i-decide-which-family-of-variance-link-functions-to-use-in-a-generalized, https://stats.stackexchange.com/questions/303514/family-in-glm-how-to-choose-the-right-one – kjetil b halvorsen Jun 10 '21 at 18:22

0 Answers0