Hi all,
I am running a trivariate Cholesky ADE model with one binary and two continuous variables. As profile-based confidence intervals took more than 24h to run and a lot of them were missing I decided to try bootstrapping. I ran mxBootstrap with 1000 replicates to get 95% CI and all looked fine except CI's for dominant genetic correlation rD. I get the following error:
> mxEval(rD, boot$MZ)
[,1] [,2] [,3]
[1,] 1.00000000 -0.56438885 -0.57273008
[2,] -0.56438885 1.00000000 -0.32630007
[3,] -0.57273008 -0.32630007 1.00000000> mxBootstrapEval(rD, boot, bq=c(0.025,0.975))
Error: The following error occurred while evaluating the subexpression 'solve(sqrt(ADE.I * ADE.D))' during the evaluation of 'rD' in model 'ADE' : Lapack routine dgesv: system is exactly singular: U[1,1] = 0
rD was defined as following:
corD <- mxAlgebra( expression=solve(sqrt(I*D))%&%D, name ="rD" )
If I am not mistaken the reason for that is the fact that the first element of D was estimated as zero. But I don't know how to get around this problem. Running 1000 replicates for this bootstrap took more than 17 hours, so I would really like to avoid rerunning it again, but I guess this is not possible?
Thank you in advance!
Julia