Fri, 05/03/2013 - 08:47

hello, Everyone!

I am a green hand for OpenMx study. Although I took a "behavior genetics" course, and have some experience about using SEM to deal with continuous data in the practical. I do not anything about liability threshold model since yesterday. Now i need estimate the heritability of two variables which both contain more than 2 categorite. I want set up two univariate liability threshold model first, and set up a bivariate model. However, there is limited information for me to start. I read some articles about liability threshold models. Almost of them are talking about dichotomous data ( like case-control study), few about ordinal data. So I am wondering is anybody here could give me as a beginner some suggestion about using liability threshold models deal with more than 2 cartegory variable data (classical review, slides or example scripts will be helpful.)? If, so I will be very appreciated!

See http://www.vipbg.vcu.edu/NIDAworkshop2012/NIDAworkshop.shtml

On Tuesday 8am, Sarah Medland gave presentation and example scripts for twin data with ordinal variables, but the session was not videorecorded.

Thank you so much Ben, I got a lot of information from the material. However, I can not find the saturated model in her script(just ACE model). I am trying to figure out how to do saturated model, but i was stuck at setting starting values.

As I know, I need three starting values.

1. Starting value for threshold. My phenotype is coded as (1,1,1,1,2,2,2,2,3,3,3,3,3) . So my starting value for threshold is 1,right?

2. Starting value for MZ correlation. I use the correlation function in SPSS, rMZ=0.952, so I need set it 0.952

3. Starting value for DZ correlation. I use the correlation function in SPSS, rDZ=0.464, so I need set it 0.464.

However, I am not sure if I am right, especially rMZ and rDZ. In SPSS I use analysis –correlation –bivariate. (twin1 ,twin2) pearson, 2-tailed, to get the correlation. Could you tell me if I did it right? but, I read somewhere Starting value for MZ correlation is expected correlation not observed so. Maybe I am still not correct here.

Thank you so much again!

Bochao Lin(the Chinese girl beside you in Dorret's dinner 7th April 2013 Boulder. :) )