Hi, all
After I did multiple imputation for missing values, I ran ANOVA test for independent variable with three group. but SPSS always remind that 'Post hoc tests are not performed for var[DV] because there are more than 50 groups'. In fact, in my dataset there are only three groups: never, sometimes, always.
I wonder why SPSS reminded me that there are more than 50 groups. thanks a lot!
Please post the command syntax for your ANOVA, and include information about all of the variables (e.g., whether they are categorical or quantitative, and how many levels there are for categorical variables).
Hello Qian,
Is it possible that you reversed your independent variable (the factor having 3 levels) and dependent variable when eliciting the SPSS analysis? This error message is consistent with that sort of mix-up.
In the oneway dialog box, the dependent variable is the first one entered.
Gppd luck with your work!
Bruce Weaver Thanks a lot! My independent variable is categorical, with three levels, and dependent variable is continuous variable.
After i checked the syntax, the problem is that i did not split file and activate the imputed dataset (see syntax information below).
Now i have two new questions about imputated dataset.
1. I wonder why there isn't pooled results when i run ANOVA. Also, the post hoc test doesn't work with imputated dataset.
2. I want to do regression analysis with a few dummy variables. Whether should i code those dummy varibles before multiple inputation or after it?
syntax infomation
Note # 5268. Command name: GET FILE
>This appears to be a Multiple Imputation dataset. In addition to the original
>data, a multiple imputation dataset contains multiple copies of the data in
>which missing values have been replaced with imputed values. If you want this
>dataset to be treated as a multiple imputation dataset, use the SPLIT FILE
>command to define the Imputation_ variable as a SPLIT FILE variable before
>performing analyses.
DATASET NAME DataSet1 WINDOW=FRONT.
ONEWAY MCPRACTICE BY language proficiency
/MISSING ANALYSIS
/POSTHOC=TUKEY ALPHA(0.05).
David Morse Thanks a lot! The problem might be that imputated dataset is not split.
I have two new questions about my imputated dataset.
1. I wonder why there isn't pooled results when i run ANOVA. Also, the post hoc test doesn't work with imputated dataset.
2. I want to do regression analysis with a few dummy variables. Whether should i code those dummy varibles before multiple inputation or after it?
thanks again!
Thanks for providing some syntax, Qian Liu. Your ONEWAY command treats variable MCPRACTICE as a quantitative dependent variable, and variables language and proficiency as categorical explanatory variables in separate one-way ANOVA models. From the help for ONEWAY:
"The analysis list consists of a list of dependent variables, keyword BY, and an independent (grouping) variable."
Is MCPRACTICE really the dependent variable, or is it the categorical explanatory variable? That could be one issue (as noted by David Morse).
Re pooled F-tests when using multiple imputation, SPSS does not provide them. See this article by Joost van Ginkel & Pieter Kroonenberg:
Article Analysis of Variance of Multiply Imputed Data
Note this bit especially:
"SPSS 19.0 does not provide any pooled results for F-tests for multiply imputed data. Using an SPSS macro (Van Ginkel, 2010a) the procedure may be carried out in SPSS but this still requires substantial manual pre-processing. For repeated-measures ANOVA the procedure is even more complicated."
HTH.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA