I'm trying to do a discriminant analysis. My data do not have normality or homocedasticity, therefore, I can not do a linear discriminant analysis, but my database is small, therefore, I can not do a quadratic discriminant analysis either. I thought that my best option is a regularized discriminant analysis, but I can not find a way to estimate the optimal values for lambda and gamma according to my database. I understand the way they influence, but I can not find a way to estimate the optimal values. Or are the values used in the author's opinion without the estimation methodology?
If you are talking about the variables selection using Wilks Lambda it available in R under the function " greedy.wilks() ".
I found the answer here:
https://daviddalpiaz.github.io/r4sl/regularized-discriminant-analysis.html
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking