-1

enter image description here

I have these two plots on a linear regression model attempt. However I can see that the normal distrubution given the QQ is violated.

Can someone give me an example of a transform that could be applied to make the distribution of the residuals closer to the normal distribution?

gauss123
  • 9
  • 1
  • Why are you interested in whether your residuals are normally distributed? – Stephan Kolassa Mar 03 '22 at 18:43
  • https://stats.stackexchange.com/q/100214/1352 – Stephan Kolassa Mar 03 '22 at 18:44
  • In order to get a model under Gauss Markov assumptions, I need to have constant variance in error terms and errors being normally distrubuted. I'm aware of that I can implement say for example cox box method. I was just wondering in a more conceptual way, how one can understand which transformation that is needed. – gauss123 Mar 03 '22 at 18:46
  • 2
    This is not a very useful piece of information for identifying a transformation. One of the better ones is a plot of the spread of the residuals against the predicted value. At https://stats.stackexchange.com/a/74594/919 I explain how to do this in a slightly different context, but it's so closely related you will have no trouble applying it. – whuber Mar 03 '22 at 20:18
  • 2
    Your qq-plot isn't labeled. Sometimes the theoretical distribution is put on the x-axis, and sometimes the y-axis. The interpretation will depend on which is the case here. – gung - Reinstate Monica Mar 03 '22 at 21:14
  • You present two graphs without saying how they were obtained and without labeling the axes. – BruceET Mar 04 '22 at 03:46

0 Answers0