|
Message/Author |
|
|
I want to use Mplus to analyze some NHIS data, a repeated cross-sectional survey (no repeated assesments on any one individual). It utilizes a multistage sample (details here: http://www.cdc.gov/nchs/nhis/about_nhis.htm#sample_design). Logistic regression (WLS) and latent class analysis will answer my questions. Because I'm looking at restricted data, I have to go through remote access to the Research Data Center and will be limted to SAS. My idea was to use SAS to generate covariance matrices for input into Mplus. Therefore, I need to figure out exactly what to feed Mplus in order to make my plan work. One analysis will focus on the entire US and another will be segregated by state. All outcomes are binary and several covariates (e.g., sex, age, race) will be included. Survey weights are provided by NHIS. I would like to compare M/F adjusting for age and race. 1) Is my plan feasible? 2) For the entire US analysis am I going to need one covaraince matrix or several? 3) Same question as in #2 for the analysis broklen down by state. |
|
|
You need raw data for these analyses. You cannot use covariance matrices. |
|
Back to top |
|
|