Dear SEM experts, We want to generate thousands of datasets with varied sample sizes (50, 200, 500) from a structural equation model (say, model with one endogenous latent variable and two exogenous latent variables, multivariate normal and skewed variables, each latent variable with three items that used 7-point ordinal options). We will use the data sets for our simulation studies. Can you please suggest a software that can generate such data sets? Eins |
This book will help you doing simulation with AMOS SEM. The built-in basic programm will help you do the required simulation.
A Beginner's Guide to Structural Equation Modeling, By Dr. Randall E. Schumacker, Dr. Richard G. Lomax |
In reply to this post by E. Bernardo
It's a matter of generating data from a variance-covariance matrix which conforms to a particular SEM, with the bells and whistles you desire (e.g., skewed variables etc.). Consequently, I would say that virtually any of the major stats software packages should be capable of simulating such data.
Ryan On Sun, Feb 24, 2013 at 1:56 AM, E. Bernardo <[hidden email]> wrote:
|
In reply to this post by E. Bernardo
I have not yet used the Simulation Builder in
SPSS, but it should get you started.
A great deal of simulation can be done using INPUT PROGRAM and the many rv.* functions. Check the archives of this list looking for "INPUT PROGRAM" and "RV." There has also been syntax posted to generate variables from a given pop correlation matrix. You should be able to adapt that syntax to create covariances. Art Kendall Social Research ConsultantsOn 2/24/2013 1:56 AM, E. Bernardo wrote:
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Art Kendall
Social Research Consultants |
In reply to this post by E. Bernardo
Dear Eins, I have done something like this in SPSS, even though I was satisfied with a few hundred datasets. Basically I wrote a CMD script for windows that contained a list of multiple parameters that were fed to a textfile, line by line, each line specifying
one set of conditions for analysis (i.e. a separate dataset). The command script started SPSS-projects. In SPSS I used python to grab these parameters, to insert them into the right places in the SPSS script
and to save the resulting datafile with a name that reveals the most important parameters used. In other words, most of my SPSS script was wrapped inside python. I am sure there are more elegant ways to do it, but this was based on my need for solid documentation for every step in the process, and the tools
I feel confident about. I am in a rush now, so I will not try to create a simplified script for example for the list, but I will send you directly a few of these files.
If someone else needs these, please let me know. Best, Eero Eero Olli phone +47 23 15 73 44 Senior Adviser at Equality- and Anti-Discrimination Ombuds office Mail: Post office box 8048 Dep, 0031 Oslo Visits: Mariboesgate 13, Oslo Dear SEM experts, We want to generate thousands of datasets with varied sample sizes (50, 200, 500) from a structural equation model (say, model with one endogenous
latent variable and two exogenous latent variables, multivariate normal and skewed variables, each latent variable with three items that used 7-point ordinal options). We will use the data sets for our simulation studies. Can you please suggest a software that can generate such data sets? Eins |
Free forum by Nabble | Edit this page |