Login

News

bivariate model (problem in saturated model)

Posted on
No user picture. JuanJMV Joined: 07/20/2016
Forums
Behavioral Genetics Models
Hi, I'm very new to openMX and I'm trying to analyse a bivariate analysis (one variable quantitative and one categorical).
But In the saturated model, there are significant differences in "Constrain expected Thresholds to be equal across twin order and zygosity" . When I remove oposite sex twins the differences between models there aren't significant.
I would like to know what model should I use.
*there are differences between men and women in the categorial variable.
A sex limitation model? or remove Oposite sex twins?
Any example or guiding light will be very appreciated.
Thanks a lot in advance!

Replied on Tue, 07/26/2016 - 16:44
Picture of user. AdminNeale Joined: 03/01/2013

I suspect that you have a large number of OS DZ twins - they should be about the same in number as the sum of the MM DZ and FF DZ pairs. Possibly it is because you have greater statistical power that you find differences when OS DZ are included. However, it may be that the OS DZs are quite different from the other twin pairs - though this would be unusual. I think a sex limitation model that allows for differences in means between males and females would likely be the best way to go.
Log in or register to post comments
Replied on Mon, 10/17/2016 - 06:38
No user picture. JuanJMV Joined: 07/20/2016

Hello again,

I hope that you still get this message.

I went for a sex-limitation model but the problem seems to be related to zygosity, not sex.

I have found that the problem arise when I try to equate the thresholds within sex and across zygosity, and not between the same-sex pairs (e.g., MZM = DZM), but between the same-sex pairs and opposite-sex individuals (e.g., (MZM=DZM) = OSDZ males). There is an statistically significant difference between OSDZ individuals and the pairs of twin of the same sex than they.

For now, I haven't found any feasible explanation for this. I'm considering several strategies for dealing with the problem:

1. Setting free the thresholds for OSDZ pairs in the genetic model, and equating the thresholds for the same-sex pairs (across twin order and zygosity).
2. Fixing all the thresholds to be equal, despite the differences found.
3. Removing OSDZ pairs from the analysis.

I haven't found any documentation supporting either of these, and I would really appreciate any insights from the OpenMx community about this.

Thanks very much again for your time.

Best regards.

Log in or register to post comments
  Login

News