

Multigroup and bias corrected bootst... 

Message/Author 


Hello: I have a multigroup SEM model that has x>y which is partially mediated by two variables M1 and M2 such that x>y and x>M1>y and x>M2>y. The grouping variable is dichotomous 0 = low and 1 = high. I am using type=complex as it is a complex sample design. I would like to estimate the bias corrected bootstrapped SE, but you can't do that with type = complex. Is there a work around for this? For example if I were to create a phantom variable and fix its path (to x) to be the product of the x>m>y paths, then remove type = complex, would I get reliable biascorrected SE for the phantom variable path since its fixed or would removing the type=complex create unreliable coefficients. 


I don't see how the phantom variable approach would correct for nonindependence of observations due to clustering but perhaps I am missing something. 

Back to top 

