We can't assign people with missing values on the grouping variable to any group so they are not included in the analysis. I'm not sure how you can change that if you don't know which group they are in.
I have reviewed other posts and still have a question about grouping variables and moderation.
I have run a multiple group model. one of the structural paths differed between the four groups as follows: group 1: (beta = .36, p less than .001), group 2 (beta = .40, p less than .001), group 3 (beta = .33, p less than .001), and group 4 (beta = .06, p = .12). Someone asked: "This shows which paths do remain significant and which do not. but it does not tell us whether the significant path DIFFERS from the nonsignificant does it?" I replied that this is correct and you could do a chi-difference test to determine statistical difference. With out this difference test, are the results of this analysis not showing moderation?
I am struggling to make the GROUPING option work. Here is my syntax:
GROUPING IS i_Mperstak_Above (0.000 = Below_Mean 1.000 = At_or_Above_Mean);
This corresponds with the User Guide, and I know that there are no missing data on i_Mperstak_Above, i.e., everyone is either 1.000 or 0.000. I created that variable in SPSS prior to bringing the data into MPlus.
The output I receive is below. Note: The model runs fine when I remove my GROUPING syntax:
*** ERROR The number of observations is 0. Check your data and format statement. Data file: c:\users\grant\box sync\michigan\dissertation files\data\grant jackson\data\all igd st
*** ERROR Note that there are observations with missing or unknown grouping values. These cases were deleted. Check that you have specified the GROUPING option correctly.
My ultimate goal is to compare how my structural equation model fits the data for these two groups using a Wald test.
I am having a similar problem as above, after using my code multiple times "GROUPING IS change (0 = nonchanger 1 = changer);"
I now get the following error:
*** ERROR Group NONCHANGER has 0 observations. *** WARNING Data set contains unknown or missing values for GROUPING, PATTERN, COHORT, CLUSTER and/or STRATIFICATION variables. Number of cases with unknown or missing values: 107
This is incorrect and I have checked my data and code multiple times. There is no missingness on this grouping variable. Thank you in advance for your help!
That sounds like you are not reading your data correctly. E.g. check that the number of columns in your data is the same as the number in the NAMES= list. If this doesn't help, send your full output and data to Support along with your license number.
I was wondering if it is possible to specify more than 1 subgroup at a time using the GROUPING command (e.g, gender, age, and treatment history)? Or would each subgroup have to be analyzed one at a time?
Hello, I have a follow-up question from the post by Patricia Roberson posted on Tuesday, June 25, 2013 - 1:58 pm and Dr. Muthen's subsequent response of "If group is the moderator, there is no moderation if the betas cannot be shown to be significantly different. At least some pairs of betas need to be significantly different."
My question: Is there an additional test to see if the betas are significantly different? If so, what test is this and how do I perform it? If no, how do I show they are significantly different.