Using a metric developed by someone else, I have calculated vegetation condition scores from a set of measured variables (about 1000 sites).

I am using simple linear regression to determine how much effect each variable (covariate) has on the score (response). I have used simple sum of squares based % deviance explained (sum of squares for the term/sum of sum squares) which works well. But I am wondering if there is a better / less biased way to estimate the effect size. Someone has suggested variance ratios (mean sq of the term/residual mean sq). However, I don't understand why these would be better or if I am calculating it right? Can anyone explain to me why variance ratios would be better. Or if there is a better effect size estimator I can use in this situation? Or, is what I am doing actually fine?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.