Are you using a fixed- or random-effects model? A complete study is only required in the fixed-effects model.
My guess is that there is not enough studies in the subset of your data.
Good afternoon,I am consulting this discussion board in regards to an error in R when running the first stage of the TSSEM approach on a subset of 6x6 matrices from a larger set. I get the following error:Error in solve.default(mxFit@output$calculatedHessian) : Lapack routine dgesv: system is exactly singular: U[1,1] = 0I have checked my matrices and none of them are non-positive definite or singular. Also, running the TSSEM1 on the full set of matrices resulted in model convergence and an…See More
"Hi Dr. Cheung, thank you for your response. I have installed the latest version of OpenMx and metaSEM, and have tried adding the model.name statement to my syntax. I am still getting the OpenMx Status 6 error.
However, I attempted to do a…"
Are you using OpenMx 2.2.6? If yes, you may try to see if it helps by including a simple character in the model.name argument. For example,
library(metaSEM)fit.error <- tssem1(Becker09$data, Becker09$n, method="REM",…"
Hi Dr. Cheung and others,I am attempting to fit a MASEM using the two-stage approach in R with metaSEM. I have 176 correlation matrices (10x10) with varying degrees of missingness, but almost all matrices (except 2) are missing at least 1 or more variables. My attempts to fit the first stage using tssem1 produced questionable results. The OpenMx status is 6 after a model using RE.type='Diag' is run, and the heterogeneity statistics for all intercepts are well over .95 (except one, I2=…See More
It is not advisable to use Mplus or other SEM packages to analyze the pooled correlation matrix. In your example, the sample sizes vary from 36 to >13,000. If one sample size is used, say the harmonic mean with 5,000, the sample sizes…"
Hi Dr. Cheung (and all), I've really enjoyed reading your new book on meta-analytic SEM!I am interested in running a meta-analysis on correlation matrices using the multivariate approach. I know one method is to do TSSEM using R, but my expertise lies in the usage of Mplus. My questions are: 1. Can I read the pooled correlation matrix calculated from the first step of TSSEM into Mplus as a summary data file? As in, is it advisable to do this?2. If this is possible, what sample size should I use…See More