

Twolevel modeling with replicate wei... 

Message/Author 


Sorry, I feel that I should know this, but can Mplus handle twolevel models and the use of replicate weights at the same time? For example, suppose the sampling design consisted of stratified selection of PSUs, then within PSUs, schools were selected and then students within schools. The replicate weights could accommodate the first level of sampling while the model addresses the 2nd and 3rd. I know that the sandwich estimator for the SEs is available if you have stratum and PSU info, but what if we only have replicate weights on the file? Thanks, Laura 


No, replicate weights are available only for TYPE=COMPLEX. 


Linda and Bengt, have there been further updates to this? How can we acknowledge the clustering of cases within higherlevel units, while using replicate weights? In my data, I have students nested in teachers/classrooms, and I have a set of replicate weights for the teachers, such that results generalize to all teachers in the U.S. The replicate weights account for the unequal probability of a teacher to be sampled, but I need to employ CLUSTER = teachID to account for the nesting of students on level 1 within teachers on level 2. The CLUSTER command is typically used with TYPE = TWOLEVEL. Is there any way to account for the nested structure of the data, while still using the replicate weights? Thank you. 


Alternatively, I have both level1 and level2 weights. I just don't have stratum and PSU identifiers for all cases. If I use TYPE=TWOLEVEL in conjunction with the WEIGHT and BWEIGHT commands to weight units on the two levels, will this adequately account for the stratification done prior to sampling, such that I don't need to specify COMPLEX on the TYPE line as well? That is, will using WEIGHT and BWEIGHT with TYPE=TWOLEVEL make it such that I do not need to specify stratum? Does using the weights for each level account for stratification? If so, is there a paper that I can read and cite for this idea? Thank you. 


Using replicate weights with type=complex accounts for the clustering and stratification (so you don't need to specify those). Here are the references http://statmodel.com/download/Resampling_Methods5.pdf Stapleton, L. M. (2008) Variance Estimation Using Replication Methods in Structural Equation Modeling With Complex Sample Data, Struc tural Equation Modeling: A Multidisciplinary Journal 15, 183  210 

Back to top 

