Page 1 of 1

compute L2 residuals from L1 and shrinkage

Posted: Thu Nov 17, 2011 4:15 pm
by gilleecl

Just another question I wonder if someone could advise on... In order to fully understand how residuals are computed, I was attempting to compute my L2 residuals by hand, i.e. by taking the average of the L1 residuals and multiplying this by a shrinkage factor. However, I don't arrive at the correct L2 residual given by MLwiN (or HLM for that matter which gives same value as MlWin). I was using the shrinkage factor as follows: sigma u sq / (sigma u2 + (sigma e2 / nj)). Could anyone advice on where I might be going wrong?

thanks alot,
Lorraine Gilleece