Confusing error message about samplin... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 Whitt Kilburn posted on Monday, April 13, 2009 - 5:15 pm
Hello,

I have a basic question about an error message regarding sampling weights: "*** ERROR Weight variable has nonpositive value at observation 64."

Counting by rows of the dataset, observation 64 does not contain a nonpositive value.

My weight variable -- only accounting for unequal probabilities of selection and for nonresponse bias -- has a minimum value of 0 and no missing values. I assume (perhaps incorrectly?) 0 should be allowed as a value for a sampling weight.

Could someone explain why I receive this error message? An excerpt from the input file for a basic 2-factor CFA is below:


WEIGHT is sweig;
ANALYSIS: TYPE = GENERAL;
ESTIMATOR = MLR;
MODEL: prob BY cert caus cons ;
poli BY plant fuel tax ;
fuel WITH tax;
OUTPUT: SAMP STAND MODINDICES ;
PLOT: TYPE IS PLOT1;

And for what it is worth, I'm a licensed user of version 3.13.

Thanks,
Whitt Kilburn
 Linda K. Muthen posted on Monday, April 13, 2009 - 6:20 pm
It sounds like you are not reading your data correctly. You can remove the weight option and put the weight variable on the USERVARIABLES list and save it using the SAVEDATA command to see what Mplus is reading.

If your data are in free format, you should check to be sure there are no blanks in the data set. Blanks are not allowed for free format data.
 Whitt Kilburn posted on Monday, April 13, 2009 - 7:43 pm
Thank you for your prompt response.

The var-cov matrix in the SAMPLE STATISTICS output matches what it should be, as do the files generated with SAVEDATA -- the cov matrix and fixed format dataset appear as they were intended to be read by Mplus.

In the output data file from SAVEDATA, the set of observations on the weight variable do not contain any missing, zero, or negative values.

I re-ran the input file, using the FIXED format dataset from the SAVEDATA command, but still received the same error message. Given that Mplus removed the observations (due to non-response on at least one variable) that had a 0 value on the weight variable, I do not understand what is causing the error.

The input file I used to produce the FIXED dataset, and verify the data, was as following:

DATA: FILE is "C:\data\gw analysis\gwdata.dat" ;
VARIABLE: NAMES ARE cert caus cons plant fuel tax sweig;
USEVARIABLES ARE cert caus cons plant fuel tax sweig;
MISSING are .;
! WEIGHT is sweig;
ANALYSIS: TYPE = GENERAL;
ESTIMATOR = MLR;
MODEL: prob BY cert caus cons ;
poli BY plant fuel tax ;
fuel WITH tax;
OUTPUT: SAMP STAND MODINDICES;
SAVEDATA:
FILE is cfav1.dat;
SAMPLE is samplecfa.dat;
PLOT: TYPE IS PLOT1;

Should I submit this question directly to support?

Thanks,
Whitt
 Linda K. Muthen posted on Monday, April 13, 2009 - 7:47 pm
We no longer support Version 3.13. The current version is 5.2. If you have explored everything, perhaps there was a problem in Version 3.13 that has now been resolved.
 Whitt Kilburn posted on Monday, April 13, 2009 - 8:16 pm
Ok, thanks. Although I had originally downloaded an upgrade from 3.13 (to version 4.x) when my support contract was up-to-date, I lost it when my hard drive crashed. I guess it is not possible for you all to verify that I did that and then allow me to upgrade to that version?

I was looking into purchasing a new support license, but I could not find how to do so under the Order Mplus Software link. I assume a new license (university single user) for the Base program could cost $125.

Do I contact support for that?

Thanks,
Whitt
 Linda K. Muthen posted on Monday, April 13, 2009 - 8:55 pm
You need to contact Mplusadmin@statmodel.com. If you give your license number, you can obtain an updated annual renewal statement.
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: