|
|
| Message/Author |
|
|
| Sophia posted on Friday, April 10, 2015 - 1:41 pm
|
|
|
I would like to run a HGLM. Observations are nested within persons (P). I have explanatory variables on Level 1 (X1, X2, X3, X4) and my outcome variable (Y) is binary. I would be grateful if you could help me with the following questions:
1) Is the difference between a multilevel model with a continuous outcome variable (HLM) and a multilevel model with a binary outcome variable (HGLM) in Mplus language that I add categorical = “my outcome variable” to the variable command?
variable: names = X1 X2 X3 X4 X5 Y P;
usevar = X1 X2 X3 X4 Y;
categorical = Y;
cluster = P;
within = X1 X2 X3 X4;
centering = grandmean (X1)
grandmean (X2)
grandmean (X3)
grandmean (X4);
analysis: type = twolevel random;
model: %within%
Beta1j | Y on X1;
Beta2j | Y on X2;
Beta3j | Y on X3;
Beta4j | Y on X4;
%between%
Y with Beta1j;
Y with Beta2j;
Y with Beta3j;
Y with Beta4j;
output: sampstat;
2)Is there a minimum cluster size (minimum of mandatory observations per person) or a minimum number of clusters in order to run the analysis (a HGLM)? Or could you recommend me a reference dealing with the necessary sample size? |
|
|
| Sophia posted on Friday, April 10, 2015 - 1:52 pm
|
|
|
| 3)Would you generally recommend to use the ML or MLR estimator for this analysis? |
|
|
|
|
Q1. Yes.
Q2. You should have at least 30-50 clusters. See the Hox multilevel book for recommendations.
Q3. MLR has some robustness built into it. |
|
|
| Back to top |
|
|
