Message/Author 

Matt C posted on Wednesday, March 17, 2010  12:21 pm



Hello, I have a fairly basic question. I ran some initial EFAs on itemlevel data that I have in SPSS. Responses to the items are reported on ordered categories. Therefore, I imported the data into Mplus (I'm still running version 3.01) using .dat files to run the analyses using EFA with categorical indicators. There are several unrelated sets of items, so I'm actually running a series of EFAs. One thing I'm noticing is that the number of observations that Mplus is reporting differs from the number of observations reported in the SPSS analyses for some of the EFAs (not all of them), despite the fact that I'm using listwise in both packages. In some cases, the number of observations reported in Mplus exceeds what I was getting in SPSS (e.g., one analysis had 697 observations used in SPSS, while Mplus reported 722). In other cases, the opposite occurs (e.g., one analysis had 560 observations used in SPSS, while Mplus reported only 43). Again, this is only occurring for certain subsets of items; in many cases, the number of observations matches between both packages. Aside from something odd that might occur when I'm importing the data, is there anything in how Mplus is handling the data that might lead to differences in the number of observations being used in the analyses? It seems somewhat unlikely, but I just wanted to Thanks for any ideas. 


I can't think of anything. It sounds like you either have blanks in your data and are reading it as free format or you are setting the data reading up incorrectly. SPSS uses blanks for missing in some cases. Blanks are not allowed with free format data. 


I have 302 observations. They are complete, no missing data. SAS can read all the 302 observations. Mplus reports only 151. 


Hello Muthens' Just to explain a little from the above post, here is the report I get  THE MODEL ESTIMATION TERMINATED NORMALLY THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NONPOSITIVE DEFINITE FIRSTORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS 0.773D16. PROBLEM INVOLVING PARAMETER 151. THIS IS MOST LIKELY DUE TO HAVING MORE PARAMETERS THAN THE SAMPLE SIZE. 


When you read only half of the observations, it means that you have more variable names in the NAMES statement than you have columns in the data set. You should not have more parameters in your model than you have observations. 


Hello! I think I have 630 data points, and in previous runs there were 630 observations reported in Mplus. Now I only get 441 observations. What is changed is that i added the following in the syntax: LISTWISE=ON; ANALYSIS: TYPE = GENERAL; MODEL = NOMEANSTRUCTURE; Everything else is the same. Why less observations? Thanks for your explanation! 


Listwise=on deletes observations that have missing data on any variable. See UG. 

Back to top 