Twin Multivariate Dataset Simulation

S. Purcell, 2000

This module is designed to

to calculate the correlation matrices resulting from a cholesky decomposition specification of multivariate twin data
to simulate datasets with properties specified as above

The Cholesky is specified as shown below (in this case for 3 variables). That is, for n observed variables, the columns represent latent variables (where the i^th latent variable acts on the i^th to the n^th observed variables) and the rows represent the observed variables.

          LV 1   LV 2   LV 3
 Trait 1  p11    
 Trait 2  p12    p22
 Trait 3  p13    p23    p33

Univariate
Bivariate
3 variables multivariate
4 variables multivariate

The simulation option produces either a page of data in comma-separated value format or space-delimited format with no headers. Either save (as text only, not as source html) or cut-and-paste into your favourite text-editor and save from there. The space-delimited format is more convenient for directly simulating data for analysis in Mx, which prefers datasets with no non-numerical symbols.

The dataset is formatted as follows :


 ID  Zyg  Twin1Trait1  Twin2Trait1 ... ... Twin1TraitN Twin2TraitN

Please note that this is not how data would usually be read into Mx. Therefore, in your script the label command should read

      Labels id zyg x1 x2 y1 y2

in the case of a bivariate analysis, where x and y are the variables for twins 1 and 2, but the select command should read

      Select x1 y1 x2 y2

Last updated 21.11.2000 by Shaun Purcell