Chi Square Test, Pre Post Study, Dichotomous Variables, Different Group Sizes

#1
Hi Everybody,

I am a medical student. I am doing a Final Year Project (FYP) and need a little help. I must admit to being a novice statistican and apologise if this question is a little dumb.

My FYP is a pre-post study (i.e. pre-intervention and post-intervention). Variables in my study for both pre- and post-intervention are dichotomous variables. I have 73 individuals in my pre-intervention group and 68 individuals in my post-intervention group. 60 individuals are paired being in both the pre- and post-intervention groups.

I have used Mc-Nemar's test to evaluate paired individuals pre- and post-intervention.

To ensure that my analysis is sufficiently rigorous I would also like to compare all 73 individuals in the pre-intervention group to all 68 individuals in the post-intervention group. I propose using the Chi-square test (Continuity correction and Linear-by-linear) for this analysis. I am using SPSS 20 for this.

Is such an approach acceptable? Is there an alternative and better approach to this analysis? Or should I just be content with my McNemar's test?

Thanks for the help,

Macho
 

Karabiner

TS Contributor
#2
Is such an approach acceptable?
No. You have paired observations. You cannot use
Chi² for independent samples in this case.
Is there an alternative and better approach to this analysis?
You could do a sensitivity analysis.
Substitute missings by values which
push your results to an extreme
(worst case).

With kind regards

K.