Error when trying to run my data but not sample data

8 replies [Last post]
dstanley4's picture
Offline
Joined: 08/07/2012

Hi Everyone,
Just trying out metaSEM. Analyzing the sample data worked perfectly. Unfortunately, the error below occurs when I try to run meta3 with my data. Note that there are no missing or NA values in my data. Any help in trying to get this to run would be greatly appreciated!
Cheers,
David

First few row of myData (unsorted):
yCol xCol country yColSamplingVar
1 0.35000 39 51 0.0020375922
2 0.48780 35 53 0.0090533427
3 0.24500 36 14 0.0040732539
4 0.41600 60 33 0.0039870241
5 0.38000 54 16 0.0016615144

My Command:
metaHLMResults= meta3(y=yCol, v=yColSamplingVar, cluster=country,x=scale(xCol, scale=FALSE), data=myData);

Error Message in R:

Error in running the mxModel:

Running Meta analysis with ML
Error in running the mxModel:

Error in runHelper(model, frontendStart, intervals, silent, suppressWarnings, :
Error: NA value for a definition variable is Not Yet Implemented.
>

**************************************************************
David Stanley, PhD
Department of Psychology
University of Guelph
Guelph, Ontario
Canada N1G 2W1

office: 4002 MacKinnon
e-mail: dstanley@uoguelph.ca
phone: (519) 824-4120, x58590
web: http://www.uoguelph.ca/iopsychology/Stanley.shtml
**************************************************************

mspiegel's picture
Offline
Joined: 07/31/2009
Oops, there's a definite bug

Oops, there's a definite bug in the identification of NA definition variables in the OpenMx 1.2 series. We'll try to issue OpenMx 1.2.5 as soon as possible.

dstanley4's picture
Offline
Joined: 08/07/2012
metaSEM Bug

Thanks Michael and Mike for your help sorting all of this out! Very much appreciated. I look forward to working with metaSEM again once OpenMx 1.2.5 is out!
Cheers,
David

mspiegel's picture
Offline
Joined: 07/31/2009
OpenMx 1.2.5 is available for

OpenMx 1.2.5 is available for download.

mspiegel's picture
Offline
Joined: 07/31/2009
We'll need more information

We'll need more information in order to determine whether there is a bug in OpenMx or metaSEM, or a mistake in the script. Can you provide the full script that is experiencing this behavior? I'm unfamiliar with metaSEM, so I'm asking whether yCol, yColSamplingVar, etc. are variables that are defined in your script. Also, we'll need access to the dataset that is causing the error. If you can't post the dataset on the forums, then you can email it to either me (mspiegel at unc dot edu) or the metaSEM developer (mikewlcheung at nus dot edu dot sg). You may find the fake data function to be useful.

Mike Cheung's picture
Offline
Joined: 10/08/2009
Your sample data works fine

Your sample data works fine though the estimates of the variances hit the lower bound (see the sample input and output files). Could you please email me the dataset and the input/output files? I will see whether it is a bug in metaSEM or OpenMx.

AttachmentSize
test.R 420 bytes
test.txt 2.49 KB
Mike Cheung's picture
Offline
Joined: 10/08/2009
David sent me some sample

David sent me some sample data. Apparently, the issue is due to the unbalanced data. If a selected subset of samples is used, it works fine.

AttachmentSize
David.pdf 123.07 KB
mspiegel's picture
Offline
Joined: 07/31/2009
We can't determine whether

We can't determine whether this is an OpenMx bug without a test case that reproduces this behavior.

Mike Cheung's picture
Offline
Joined: 10/08/2009
Sorry. The followings are the

Sorry. The followings are the syntax and the data.

AttachmentSize
SampleData.csv 890 bytes
David.R 434 bytes