0

Hello I would like to know, whether my residual vs fitted plot is okay ie does not violate the normality assumption residual vs fitted

The design is a random block design

as well as the histogram seems to be binomial or an F. can someone recommend a transformation i can use so that i can not violate the anova assumptionsenter image description here

Mason
  • 1
  • 1
    Before you spend too much time worrying about normality of residuals you might like to read https://stats.stackexchange.com/questions/29731/regression-when-the-ols-residuals-are-not-normally-distributed which explains the issues. – mdewey Oct 14 '18 at 14:35
  • 2
    Is your outcome variable continuous ? The plot indicates that it might not be and you might need to fit a (maybe poisson) glm – Robert Long Oct 14 '18 at 15:36
  • 1
    Your response is a count, right? – Glen_b Oct 14 '18 at 23:41

1 Answers1

0

The first graph indicates that variance of residuals increases along the fitted value. So the first thing is to find an transformation (maybe log transformation) such that the variance being a constant. Then check the normality.

user158565
  • 7,032
  • 2
  • 9
  • 19
  • Umm, the plot rather indicates a discrete response variable (judging from the variable name probably a count). OP should use a GLM. – Roland Oct 17 '18 at 06:48