Message/Author |
|
Anonymous posted on Thursday, March 28, 2019 - 11:32 am
|
|
|
Hello, I am running a multi-level (2-2-1) mediation model. I would like to utilize the monte carlo bootstrapped confidence intervals for estimating the indirect effects of the model. I'm having trouble understanding how to enter the commands in order to get this. Here is my syntax for the model I have run if it helps USEVARIABLES ARE group m controlv y x; BETWEEN ARE m controlv x; CLUSTER IS group; ANALYSIS: TYPE IS TWOLEVEL RANDOM; MODEL: %WITHIN% y; %BETWEEN% y; m ON x(a); y ON m(b); y ON x(c); y m ON controlv; MODEL CONSTRAINT: NEW(indb); indb=a*b; OUTPUT: CINTERVAL; TECH1 TECH3 TECH8; Thank you! |
|
|
Bootstrap for MonteCarlo simulations works in recent Mplus versions. Use Bootstrap = 500; in the Analysis command and request Cinterval(bootstrap) in the Output command. |
|
Anonymous posted on Wednesday, June 03, 2020 - 2:55 am
|
|
|
I run a 3-level model and would like to obtain likelihood-based confidence intervals. To my understanding, the default are Wald CIs. How do I get LB CIs instead? Thank you very much! |
|
|
It is not implemented in Mplus but you could do it manually. Fix all the parameters to the estimated values. To get the CI for one parameter A you must find the values A+x1 and A-x2 for which the likelihood is L0-3.84/2, where L0 is the likelihood from the original run. You would be running multiple runs with just 1 iteration to extract the log-likelihood at various values of A, i.e., you would fix that parameter at various values. The Wald CI would provide good starting values. See http://psych.colorado.edu/~carey/pdffiles/confidenceintervals_neale.pdf |
|
Back to top |