Ex. 9.3 random intercepts in twolev... PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 student07 posted on Monday, July 30, 2007 - 10:36 am
Hi,

my question refers to Example 9.3 in the Mplus userguide.

The Between-part of the model shows a correlation between the random
intercepts of y1 and y2. In principle, however, could one also regress the between-variance of y2 (= the random intercept) on the between-variance of y1?

Many thanks!
 Linda K. Muthen posted on Monday, July 30, 2007 - 12:08 pm
Instead of regressing y1 and y2 on w, you could regress y1 on y2 and w or y2 on y1 and w.
 Jeremy Flaherty posted on Wednesday, August 08, 2007 - 1:33 pm
A couple more questions based on Example 9.3 still:

1) In the between part, could y1 be regressed on w, as well as y2 on y1 and w?

%BETWEEN%
y1 ON w;
y2 ON y1 w;

2) If so, does this apply to situations where the dependent variables are both categorical or where one is categorical and one a factor?

%BETWEEN%
u1 ON w;
u2 ON u1 w;

or

%BETWEEN%
fb BY u2-u3;
u1 ON w;
fb ON u1 w;

Thanks.
 Linda K. Muthen posted on Tuesday, August 14, 2007 - 2:08 pm
1. You can regress y1 ON w.
2. Yes, dependent variables can be categorical and the can also be latent vriables.
 Jeremy Flaherty posted on Friday, August 17, 2007 - 5:21 pm
With the below input, I get two errors:
*** ERROR in Model command
Observed variable on the right-hand side of a between-level ON statement
must be a BETWEEN variable. Problem with: u1
*** ERROR
The following MODEL statements are ignored:
* Statements in the BETWEEN level:
y1 ON u1


CATEGORICAL = u1 u2;
CLUSTER = cluster;
WITHIN = x1 x2 x3 x4;
BETWEEN = x5 x6 x7;
ANALYSIS:
TYPE = TWOLEVEL;
ESTIMATOR = MLR;
INTEGRATION = MONTECARLO;
MODEL:
%WITHIN%
u1 u2 ON x1 x2 x3 x4;
y1 ON x1 x2 x3 x4 u1 u2 ;
%BETWEEN%
u1 u2 ON x5 x6 x7;
y1 ON u1 x5 x6 x7;

When I treat u1 as continuous, I get the same error, but when I treat both u1 and u2 as continuous, I get no errors and the model runs. Am I doing something wrong? Or will Mplus not do this?

Thanks
 Bengt O. Muthen posted on Saturday, August 18, 2007 - 3:43 pm
It is not directly achieved in Mplus, but it is possible to do it by creating a factor ej behind uj on the between level

%BETWEEN%
e1 by u1; e2 by u2;
e1 e2 ON x5 x6 x7; y1 ON e1 x5 x6 x7;
 Emily Blood posted on Friday, June 17, 2011 - 11:49 am
Does the following fit a logistic model at the WITHIN level and a linear model at the BETWEEN level? I thought it did (based on Ex. 9.3), but in the BETWEEN output, I get a "threshold" for condom, not an intercept, and I get no "threshold" or intercept at the WITHIN level. Note, I also get a residual variance of 'condom' at the between level which I wouldn't think I would get if fitting a logistic.
CATEGORICAL ARE condom partdid;
CLUSTER=ID;
BETWEEN = age gender enis bdi;
WITHIN = naf sweekend partdid;
ANALYSIS:
TYPE = TWOLEVEL ;
MODEL:
%WITHIN%
condom on naf sweekend partdid;
%BETWEEN%
condom on age gender bdi enis;
 Linda K. Muthen posted on Friday, June 17, 2011 - 1:33 pm
Following is the multilevel random intercept model:

yij = γ00 + γ01 wj + u0j + β1 xij + rij

The terms in order are: intercept (minus threshold), between-level regression coefficient, between-level residual, within-level regression coefficient, and within-level residual.
 Emily Blood posted on Friday, June 17, 2011 - 1:46 pm
I'm sorry, but I think some symbols did not come through in your post. The equation given is a series of &'s, #'s and 946's and 947's. I have tried two browsers.
Could you please let me know whether the within model is logistic (since the DV is the dichotomous observed variable) and between model is linear (since DV is the continuous random individual intercept). This is not clear from the output since a threshold is given in the between output which usually only happens when a non-linear model is fit.
 Linda K. Muthen posted on Friday, June 17, 2011 - 2:54 pm
I'm sorry. I didn't notice that. See slide 49 of the Topic 7 course handout. The formula is there. The within coefficient is logistic and the between coefficient is linear. There is only one threshold. If you see the slide, you should understand.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: