Background:

We are looking at assessing the effect a series of optional tutorials delivered during a semester has on test scores in a cohort of students. There will be a total of 9 tutorials delivered, attendance will be recorded. As students will be allowed to attend of their own free will, strict control vs intervention groups will not exist (we cant say 1/2 of cohort must attend all 9 and other 1/2 cannot attend any sessions). Instead, a minimum attendance (e.g 2 of 9 sessions, 3 of 9 sessions, this has yet to be decided), will be chosen as a threshold for entry into the intervention group. We may later do an 'intention to treat' analysis to see if greater attendance correlates with higher scores for within group analysis.

Methods:

50 Students were given a 20 MCQ exam, 1 mark per question answered correctly. They were allocated to either Paper A or Paper B (both 20 MCQ each), 25 students per paper. Testing was mandatory for all students to attend (50 of 50 attended). Results were as follows

Paper 1. MEAN: 11.32 MED: 11 SD: 2.39

Paper 2. MEAN: 10.56 MED: 10 SD: 2.75

(p=0.303)

Then ran a few tests to check for equal variances, difference between means and if the data followed a normal distribution, all were not significant thankfully

Variance - F Test

Difference between means - T Test

Data follows normal (Gaussian) distribution -D'Agostino-Pearson omnibus test

The 2nd test period is due to occur in June. All students given Paper A will be given Paper B and vice-versa (cross over).

So now i am wondering....

If i am to measure the test score difference between the intervention group (e.g. attended minimum 33% of sessions), and the control group (attended < 33% or no sessions), can i use an ANCOVA if the pre-test scores come from separate tests (A and B) that have been deemed to not be statistically significantly different from one another (p=0.303)? Would this be the best test to use?

Any help would be much appreciated!