Plausible values and complex survey design
Posted: Sun Jun 01, 2014 3:56 pm
Hi Chris et al.
Is it possible to advocate for additions and prioritization for your (probably big) to-do list?
1. First and foremost, which should not be too difficult to implement: Selecting variables which are a set of plausible values (this exists in PISA, TIMSS, PIRLS, etc). The quick and dirty solution would be to clarify how the Master imputation file is specified, as in Mplus, so the user could do it her/himself. But that would be unnecessarily clumsy in my opinion.
2. Complex survey stratification: Sometimes using the stratification variable in a fixed effect transformation for model-based inference is just impossible with hundreds or more strata. Mplus incorporate stratification as a an almost invisble variable (it's magic, I'm not asking what's going on behind the curtain!).
Cheers
Stephan
Is it possible to advocate for additions and prioritization for your (probably big) to-do list?
1. First and foremost, which should not be too difficult to implement: Selecting variables which are a set of plausible values (this exists in PISA, TIMSS, PIRLS, etc). The quick and dirty solution would be to clarify how the Master imputation file is specified, as in Mplus, so the user could do it her/himself. But that would be unnecessarily clumsy in my opinion.
2. Complex survey stratification: Sometimes using the stratification variable in a fixed effect transformation for model-based inference is just impossible with hundreds or more strata. Mplus incorporate stratification as a an almost invisble variable (it's magic, I'm not asking what's going on behind the curtain!).
Cheers
Stephan