Page 1 of 1

constraining error variance?

Posted: Mon Jul 19, 2010 3:30 pm
by joaordaniel
I've been unsuccessfully trying to replicate the following
"[...] The program MLwiN (Goldstein et al., 1998) was used to estimate model (θm = μθ + Um + Em) by iterated weighed least squares. Similarly to the procedure of modeling a meta-analysis with known level-1-variation (e.g., Bryk and Raudenbush, 1992; Goldstein, 1995), this was done by creating a pseudo-level in such a way that each unit at level 2 has only one unit at level 1. The error variance was constrained to be equal to the squared standard error, and the true variance was modeled at level 2." (Lubbers & Snijders, 2007).

How can I constrain the error variance (Em)?