I conducted two types of test before and after experimental process; A and B. Since I am testing the effect of tests on two groups; Control and Treatment. The original sample size was 100 i-e; 50 in each group. Later on, the size was reduced to 88 i-e; 44 per group. Since the gender distribution in test A between two groups was 88(44 in control group and also in experimental group with 22 males and 22 females ). Similar is the case with B except the gender distribution is unequal; 26 males and 18 females in control group and 22 males and 22 females in Experimental group.
Since the size is equal i-e; 88; 44 per group for both type of tests but the gender distribution is unequal in second test. This will gain the attention of audience. How come 22 males increase to 26 males? How can I defend this? Or is there an alternative solution without shrinking the sample size of 88?
If I understand correctly, there are two approaches:
1) If you are pairing the answers before and the answers after --- (That is, like a repeated measures analysis) --- then you need the same individual people in each group.
2) If you aren't pairing the answers before and after, then having a different number of males and females probably doesn't matter at all. But you can include the effect of sex in the model, so it is like a two-way anova, IV's are Time and Sex.
I recommend approach 1, even though it will reduce your sample size.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size