Clear Filters
Clear Filters

What formula for r-squared does fitlm use in the weighted linear regression case?

10 views (last 30 days)
If I use fitlm as:
mdl = fitlm(X,Y,'Weights',W);
can someone tell me the formula used in computing the r-squared that's here:
mdl.Rsquared.Ordinary
I am seeing a different result when compared to JMP. Thanks.

Answers (1)

Matt J
Matt J on 7 Feb 2024
  2 Comments
Paul Safier
Paul Safier on 7 Feb 2024
Edited: Paul Safier on 7 Feb 2024
Thanks @Matt J. It says Rsq = 1 - SSE/SST. Not seeing where the equation for SSE or SST is given in the weighted LS case. It ought to be different for OLS vs WLS. I'm interested in what MATHWORKS uses specifically for WLS.
Matt J
Matt J on 7 Feb 2024
Edited: Matt J on 7 Feb 2024
I don't believe weights are taken into account in the computation of R2. This should be easy to verify, by comparing the stack exchange formulas (both weighted and unweighted) to what Matlab gves you.

Sign in to comment.

Products


Release

R2023b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!