I am trying to run my data through SPSS GLM/Repeated Measures. My Within-Subjects Factor (dependent variable) requires 12 levels. My Between-Subjects (independent variable) is a continuous variable. I have 7 variables entered as covariates (two surveys at T0 and T1, a dichotomous (yes/no) variable, age (grouped into 8 categories), and gender . Up until the 6th covariate everything goes well. When I enter the 7th covariate and run the analysis I loose the F statistic values and significance from my Tests of Within-Subjects Effects table? I understand that the more covariates I enter the greater stress I place on the model, but not sure how to handle gender, the 7th variable? My sample is small (only 65-participants), and majority were female (only 17% male). Even with age entered into the model the F statistic remains significant where applicable? Should I keep gender out of the model?
I suspect that your model is overparametrized. Did you perform exploratory data analysis to examine the relationships between your predictor variables and covariates? You might find that most males are included in a single level of another covariate or predictor variable. In that case, sex minimally contributes to explain the residual variation when that variable is included in the model. If you don't actually have an hypothesis regarding the effect of sex on the outcome variable, you may consider dropping it from the analysis.
I haven't tried an exploratory data analysis. You are probably right about the males. I will take another look at the data in SPSS tomorrow. Thank you for your response! It was very helpful.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA