Missing data in moderated mediation m... PreviousNext
Mplus Discussion > Missing Data Modeling >
Message/Author
 Nicole Kratky posted on Monday, September 05, 2016 - 8:00 am
Dear Linda and Bengt,

I am trying to run a moderated mediation model with a dichotomous moderator (w). For this I am using the following code:

DEFINE:
XW = x*w;

ANALYSIS:
TYPE = GENERAL;
ESTIMATOR = ML;
BOOTSTRAP = 10000;

MODEL:
[x];
[y] (b0);
y ON m (b1);
y ON x (cdash);

[m] (a0);
m ON x (a1);
m ON w (a2);
m ON XW (a3);

MODEL CONSTRAINT:
NEW(LOW_W HIGH_W
IND_LOWW IND_HIW
TOT_LOWW TOT_HIW);

LOW_W = 0;
HIGH_W = 1;
IND_LOWW = a1*b1 + a3*b1*LOW_W;
IND_HIW = a1*b1 + a3*b1*HIGH_W;
TOT_LOWW = IND_LOWW + cdash;
TOT_HIW = IND_HIW + cdash;

Model is running using this, but I have to questions:

1. I have missing data on the x variable which I'd like to impute. Is there any option to do so? In the simple mediation model I fixed this by adding [x]; to the model command. But here it is not working doing so.

2. I would like to add a second x variable into the mediation model, while the moderator is only influencing the path x1-m not the path x2-m. Is this possible?
Thanks a lot!
 Bengt O. Muthen posted on Monday, September 05, 2016 - 3:58 pm
1. I think you have to also mention the mean of the XW variable.

2. Yes, this is straightforward.
 Nicole Kratky posted on Monday, September 05, 2016 - 11:02 pm
1. It's running, thanks!

2. Am I doing it this way:
MODEL:
[y] (b0);
y ON m (b1);

y ON x1 (c1dash);
y ON x2 (c2dash);

[m] (a0);
m ON x1 (a1);
m ON x2 (a4);
m ON w (a2);
m ON XW (a3);

MODEL CONSTRAINT:
NEW(LOW_W HIGH_W
IND_LOWW IND_HIW IND_x2
TOT_LOWW TOT_HIW TOT_x2);

LOW_W = 0;
HIGH_W = 1;

IND_LOWW = a1*b1 + a3*b1*LOW_W;
IND_HIW = a1*b1 + a3*b1*HIGH_W;
IND_x2 = a4*b1;

TOT_LOWW = IND_LOWW + c1dash;
TOT_HIW = IND_HIW + c1dash;
TOT_x2 = IND_x2 + c2dash;

It's running but I am not sure if I get the indirect effects right?
 Bengt O. Muthen posted on Tuesday, September 06, 2016 - 3:30 pm
Please send your full output and license number to Support.
 Mark Burton posted on Friday, September 09, 2016 - 12:06 pm
Could you explain why adding the mean of a variable to the model syntax allows for analyses of the entire sample? How is it imputing missing data? For example, pertaining to the previous post, why does adding [X] to the model command "fix" this problem. I am running similar analyses and want to make sure I am not simply doing a mean replacement or something. Thank you!
 Bengt O. Muthen posted on Friday, September 09, 2016 - 3:38 pm
Mentioning the mean or variance of an x variable brings it into the model, that is, it is no longer conditioned on but treated like the y variables. The theory for this is explained in our new book (see our homepage). No data imputation is done but simply ML estimation under MAR - which sometimes called FIML.
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: