www.cmm.bristol.ac.uk/forum

Hi

Can anyone point me in the direction of the equation used by mlwin to calculate standard errors for level 3 residuals?

Level 2 residuals seem to have the following standard errors

SQRT { (B*W) / (B*n + W) }

where B is the level 2 variance and W the level 1 variance and n in the number of level 1 units.

However, I'm not sure how to adapt this when there is a third level.

Thanks!

The residual estimation method should correspond to Appendix 2.2 of Professor Harvey Goldstein's Multilevel Modelling book (second edition available at http://www.bris.ac.uk/cmm/team/hg/multbook1995.pdf). The formula for estimation of the covariance matrix of the residuals should be equivalent to equation 2.2.3.