|
|
Discrepancies with ALGORITHM=INTEGRATION |
|
Message/Author |
|
Ian Cero posted on Thursday, February 20, 2014 - 10:13 am
|
|
|
Hello Drs. Muthen, I recently finished double-checking some output that involved missing data and found some odd discrepancies between the results. I am wondering if you can help me understand what is going on in these analyses. Both analyses are regressions using MLR. There are several predictors, including quadratic and interaction effects, but only one dependent variable. Syntax is identical in each case, except for the commands that involve missing data (i.e., Auxiliary = (m)...; [variable 1, variable 2,...]; and ALGORITHM=INTEGRATION), which are either included or removed depending on the output. In the first output, when missing data are ignored and list-wise deletion is utilized, I get sample statistics that are highly convergent with what I see in SPSS; however, when I use auxiliary variables and ALGORITHM=INTEGRATION, my sample statistics (and slopes and p-values) all change substantially. Can you help me understand what has happened here? Have I made some kind of error? Some of my results shown are below. -Ian |
|
Ian Cero posted on Thursday, February 20, 2014 - 10:14 am
|
|
|
Ooops! Here are the results I promised above: ----SPSS Results------ SPSS Mean 1 = 7.017 SPSS Mean 2 = 11.797 SPSS Correlation(1,2) = 0.607 ----Listwise Deleted Mplus Results---- Mplus Mean 1 = 7.415 Mplus Mean 2 = 11.395 Mplus Correlation(1,2) = 0.599 ---Mplus Results with INTEGRATION--- Mplus Mean 1 = 150.581 Mplus Mean 2 = 164.948 Mplus Correlation(1,2) = 0.260 |
|
|
Please send the Mplus outputs and your license number to support@statmodel.com. Please limit posts to one window. |
|
Back to top |
|
|