2 Monte Carlo Questions PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Marcus Butts posted on Saturday, July 23, 2005 - 12:12 am
I'm trying to run a monte carlo simulation in which I have two different population data sets (with X # of repititions) and then I 'test' two stacked models using those population matrices.

#1 Can I get Mplus to generate two different population data sets (I know I can model two groups)?

#2 The output doesn't provide within group standardized coefficients (or any standardized coefficients for that matter). Which is problematic b/c I want to check my original simulation paramaterizations. Is there any way to get standardized information? Or can I just simulate the data and then use it as input in a stacked model and ask for standardized output?
 Linda K. Muthen posted on Saturday, July 23, 2005 - 10:31 pm
1. Multiple groups are multiple populations. I'm not sure what distinction you are making here.

2. Monte Carlo does not provide standardized parameter estimates. I'm not sure why this would help you check your original simulation parameterizations. You generate data based on certain parameter values in Mplus and the Monte Carlo results are compared to that.
 Marcus Butts posted on Sunday, July 24, 2005 - 10:21 pm
Please disregard my first question. I figured that out (see below). However, I still have a couple more questions if you don't mind.

As a little backgroud, let me tell you briefly that I'm working with my MP (Bob Vandenberg - who sends his regards) to try to generate two random populations based on various conditions of reliability and factor loadings (and each population may be different from one another depending upon condition). Then, we test a stacked model for the two groups (calling the appropriate population matrix per group). All of this can be seen below.

MONTECARLO:
NAMES ARE x1-x12;
NGROUPS = 2;
NOBSERVATIONS = 500 500;
NREPS = 20;
SEED = 12345;
RESULTS = results2.txt;
MODEL POPULATION:
x1-x12@.631575;
LV1 BY x1@.60698 x2@.60698 x3@.60698
x4@.60698;
LV2 BY x5@.60698 x6@.60698 x7@.60698
x8@.60698;
LV3 BY x9@.60698 x10@.60698 x11@.60698
x12@.60698;

LV1 WITH LV2@.3 LV3@.3;
LV2 WITH LV3@.3;

LV1@1.0;
LV2@1.0;
LV3@1.0;

MODEL POPULATION-g2:
x1-x12@.631575;
LV1 BY x1@.60698 x2@.60698 x3@.60698
x4@.60698;
LV2 BY x5@.60698 x6@.60698 x7@.60698
x8@.60698;
LV3 BY x9@.60698 x10@.60698 x11@.60698
x12@.60698;

LV1 WITH LV2@.3 LV3@.3;
LV2 WITH LV3@.3;

LV1@1.0;
LV2@1.0;
LV3@1.0;
MODEL:
LV1 BY x1@1.0 x2 x3 x4;
LV2 BY x5@1.0 x6 x7 x8;
LV3 BY x9@1.0 x10 x11 X12;

LV1 WITH LV2 LV3;
LV2 WITH LV3;
MODEL g2:
LV1 BY x1@1.0 x2 x3 x4;
LV2 BY x5@1.0 x6 x7 x8;
LV3 BY x9@1.0 x10 x11 X12;

LV1 WITH LV2 LV3;
LV2 WITH LV3;
OUTPUT: TECH9;

We're interested in the change in fit statistics across conditions (e.g., chi-sqr diff test). However, we first want to make sure we are setting our simulation up correctly. As you can see from above, I simulated identical estimates for the two populations and I tested identical models (that should have fit the population almost perfectly - within chance). The fit indicies supported that contention. However, as per the results below in selected output, why is my 95% cover not .95 (or higher)for all estimates? Secondly, why are my population estimates 1? Where does this number come from? Similarly, are my average parameter estimates unstandardized coefficients (if not, what are they)? Also, can you clarify for me what the % significant column represents?


Group G1
Pop Avg 95% cov % sig
LV1 BY
X1 1 1.000 1.00 0.00
X2 1 0.995 0.90 1.00
X3 1 0.993 0.95 1.00
X4 1 1.002 0.85 1.00

LV2 BY
X5 1 1.000 1.00 0.00
X6 1 1.032 1.00 1.00
X7 1 1.045 0.95 1.00
X8 1 1.020 0.95 1.00

LV3 BY
X9 1 1.000 1.00 0.00
X10 1 1.019 0.95 1.00
X11 1 1.030 0.95 1.00
X12 1 1.004 0.95 1.00

LV1 WITH
LV2 0 0.101 0.05 0.95
LV3 0 0.110 0.00 1.00

LV2 WITH
LV3 0 0.102 0.00 1.00

Variances
LV1 0.05 0.382 0.00 1.00
LV2 0.05 0.348 0.00 1.00
LV3 0.05 0.377 0.00 1.00

Residual Variances
X1 0.5 0.637 0.15 1.00
X2 0.5 0.624 0.25 1.00
X3 0.5 0.628 0.40 1.00
X4 0.5 0.641 0.20 1.00
X5 0.5 0.636 0.10 1.00
X6 0.5 0.625 0.35 1.00
X7 0.5 0.633 0.35 1.00
X8 0.5 0.634 0.40 1.00
X9 0.5 0.621 0.40 1.00
X10 0.5 0.620 0.35 1.00
X11 0.5 0.624 0.35 1.00
X12 0.5 0.651 0.15 1.00

Group G2

LV1 BY
X1 1 1.000 1.00 0.00
X2 1 0.992 0.95 1.00
X3 1 0.951 0.95 1.00
X4 1 0.995 1.00 1.00

LV2 BY
X5 1 1.000 1.00 0.00
X6 1 0.970 0.90 1.00
X7 1 1.014 0.90 1.00
X8 1 1.002 1.00 1.00

LV3 BY
X9 1 1.000 1.00 0.00
X10 1 0.948 1.00 1.00
X11 1 1.007 1.00 1.00
X12 1 1.001 0.95 1.00

LV1 WITH
LV2 0 0.122 0.00 1.00
LV3 0 0.113 0.05 0.95

LV2 WITH
LV3 0 0.108 0.00 1.00

Variances
LV1 0.05 0.399 0.00 1.00
LV2 0.05 0.374 0.00 1.00
LV3 0.05 0.376 0.00 1.00

Residual Variances
X1 0.5 0.621 0.30 1.00
X2 0.5 0.623 0.35 1.00
X3 0.5 0.653 0.05 1.00
X4 0.5 0.634 0.25 1.00
X5 0.5 0.618 0.45 1.00
X6 0.5 0.646 0.20 1.00
X7 0.5 0.637 0.35 1.00
X8 0.5 0.619 0.30 1.00
X9 0.5 0.628 0.30 1.00
X10 0.5 0.629 0.30 1.00
X11 0.5 0.643 0.20 1.00
X12 0.5 0.628 0.35 1.00

And lastly, I understand how I can save each replication in a seperate results file. However, we intend to have 1200 replications (we only ran 20 this time as a test). Is there anyway to have mplus output all replications into one file? If not, when running a model is there a way for mplus to read all 1200 replications without listing them all?

Sorry for all the questions at once. (Bob and) I truly appreciate your time and consideration and have always been pleased with mplus as well as the technical help we get from you all. Thanks in advance.

Marcus
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: