SE MultiLevel Models

hlsmith

Omega Contributor
#1
Not accounting for dependence in observations and not using the appropriate cross-sectional multilevel model results in parameter estimate's standard errors (SE) being biased downward. WHY?
 

hlsmith

Omega Contributor
#2
Is this because if you did use a Multi-level model, that you now would have an additional error term. With this additional term the prior SE from a one level model would now account for a smaller proportion of the new total variation?

And/or

Does it deal with misspecification of the model leaves terms out, in which the appropriate model would require a larger sample size due to more terms?