| Attachment | Size |
|---|---|
 1saturated no siblings (running ok).txt | 2.46 KB |
| 2saturated with siblings (showing the error).txt | 2.96 KB |
Hello everyone,
I have a saturated model for categorical data with 4 categories (3 thresholds) the model is running ok for twin 1 and twin 2, but when I try to add siblings into the model I get the following error:
Error: In model 'baseModel' I was expecting 3 thresholds in column 'vars1' of matrix/algebra 'MZf.expThreshfMZ' but I hit NA values after only 0 thresholds. You need to increase the number of thresholds for 'vars1' and give them values other than NA
I am adding 2 siblings, so I have added 6 more thresholds corresponding to sibling 1 and sibling 2, I don't exactly understand what I am doing wrong.
I attached the codes below
file 1 saturated no siblings (running ok)
fil 2 saturated with siblings (showing the error)
Any enlighten from someone would be very much appreciated!
PS: I am new in both openmx and twin modelling.
Printer-friendly version
Hi
I can't reproduce your results because the nndata object is not included. I can see that the threshMf looks ok (thresholds ordered and there are 4 columns etc), but perhaps you need a threshnames= argument to mxExpectationNormal to tell it which are the ordinal variables? Just guessing here... Did the no siblings case actually estimate thresholds etc and everything fine?
What do you see if you run
mxVersion()at the R prompt? Also, how much missing data is there?This issue has been reported previously, in threads 2953 and 3996. I think it reflects a bug that has never been repaired, because none of us developers have ever been able to reproduce it, because we have lacked access to datasets in which it appears. There is now function
mxGenerateData()that can randomly generates data similar to yours. Could you generate a random dataset and post it?Also, in the past, a workaround for this issue was apparently to delete all twin pairs in which any missing data occurred. If you drop rows with missing data from your dataset and try to run the model (assuming you still have enough data to fit the model), what happens?