You are here

Violation of assumption of equal variances MZ and DZ

2 posts / 0 new
Last post
koak's picture
Joined: 05/10/2012 - 23:13
Violation of assumption of equal variances MZ and DZ


My data of cognitive measures is indicating a number of violations of the assumptions of the twin model. I have transformed the data using z scores and still find that MZ and DZ variances are different. Any ideas on how I should proceed, including how to adjust the script in the ACE/ADE model to deal with this?

Many thanks


neale's picture
Joined: 07/31/2009 - 15:14
Careful with the measures, but here's how

Essentially, you take the twin expectations and pre & post multiply by a matrix with causal paths between the twins in it. So if you had

# Algebra for expected variance/covariance matrix in MZ
expCovMZ <- mxAlgebra(name = "expCovMZ",
                      expression = rbind (cbind(A+C+E, A+C),
                                          cbind(A+C,   A+C+E)))

It might become

matI  <- mxMatrix(name= "I", type="Iden", nrow = nv, ncol = nv)
matB <- mxMatrix(name= "B", type="Symm", free=c(F,T,T,F), labels=c(NA,b,b,NA), nrow = nv, ncol = nv)
# Algebra for expected Mean and Variance/Covariance Matrices in MZ & DZ twins
expCovMZ <- mxAlgebra(name = "expCovMZ",
                      expression = solve(I-B) %&% (rbind (cbind(A+C+E, A+C),
                                                          cbind(A+C,   A+C+E))))

You'd need to include the matI and matB objects in the mxModel() of course. And change the expCovDZ formula in exactly the same way (with the same I and B matrices).