Issue with High Variance of Count Data PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Rudradev Sengupta posted on Wednesday, March 01, 2017 - 2:47 am
Hi,

I am a new user of Mplus. I have Count data (countvar) as one of the 3 variables involved in the model. But the variance of the count variable is quite large and also consists many 0s. When I transform them, for example to log-scale then my code runs perfectly and produces similar results as SAS. But when I use them as it is, by specifying:

COUNT IS COUNTVAR(i);

I get the following error messgae:

One or more variables have a variance greater than the maximum allowed of 1000000.
Check your data and format statement or rescale the variable(s) using the DEFINE command.

Is there some other way to allow for higher variance of the count data or specify them as zero-inflated data ?
Any help is much appreciated.
Thanking you in advance.

Regards,
Rudradev Sengupta
 Bengt O. Muthen posted on Wednesday, March 01, 2017 - 8:00 am
Check your data to see if you have some extremely large counts - perhaps they are due to coding error or subjects who are outliers and could be deleted.
 Rudradev Sengupta posted on Thursday, March 02, 2017 - 4:17 am
Hi,

Thanks for your reply. My data consists of some very high counts and many 0 counts, but that is the correct format of the data. There is no coding error and also I need the high counts in my analysis. Is there a way to increase the maximum allowed variance ?

Regards,
Rudradev
 Bengt O. Muthen posted on Thursday, March 02, 2017 - 5:51 pm
Our stoppage criterion (variance) is very liberal so I wonder how many high counts you have and how high they are? If you have a small subgroup with much higher counts than the others, then maybe you should analyze them separately.
 Aurelie Lange posted on Friday, April 19, 2019 - 2:07 am
Dear Dr Muthen,

I also have a question regarding the variance of one of my variables. I have two variables, measured at t1 and t2, ranging from 0 - 65. The variables are quite evenly distributed over the whole range. Surprisingly, however, the variance in my output is > 200. The intercept is 18 for t1 and -18 for t2.
These observations make me suspicious of the outcomes of the model. May something have gone wrong?

Sincerely,
Aurelie
 Bengt O. Muthen posted on Friday, April 19, 2019 - 1:49 pm
Hard to say without seeing the full output. Intercepts depend on the values of the predictors. Regarding variances, perhaps you have some outliers that are incorrectly coded or you did not give the correct missing data flag. If this doesn't help, send full output to Support along with your license number.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: