How to calculate sample size and effect size for comparing means from two groups

[Damo123]

Hi all, hoping i can find some answers here for a bit of a problem i am having regarding a sample size calculation for a study I am soon to conduct.

The population size is 64 and fixed. 32 control and 32 intervention. No drop outs will occur in this trial. There will be a pre (baseline) test on all subjects (64) and post on each of the 2 groups of 32. The test will assess knowledge in the form of multiple choice questions. Each question will be assigned a mark of 1. I seek to show a statistically significant difference between both groups post intervention. My alpha is set to 0.05 and beta at 0.20. The two groups are identical in terms of selection of subjects (same age, gender, etc).

Ideally we wish to show that in the intervention group there is a greater change in mean scores when compared to control group. I wish to detect a difference of at least 5% between the two groups.

What i was wondering is:

- What is my required sample size for each group? As each group of participants is set at 32 and will not change, can i merely > the number of questions administered until reach an ideal number? I was estimating around 20 questions, 1 mark per question so 32X20X1 = 640 for each group.

Also how does 'effect size' come into play, i have heard this term but lack the statistical knowledge to understand it.

Much appreciated to anyone who can show me the working out for this

Damian

 

[hlsmith]

The thing you need to think about, what do you hypothesize the mean group scores will be and what their standard deviations be (e.g., group #1, mean 15 and stdev 3 versus group #2 mean 13 and stdev 4). These are pieces of information needed in calculation of the sample size.

 

[Damo123]

Hi guys,

So assume mean of 11/20 and SD of 2. The baseline test involves all the same subjects (n=64) and the SD would be 2. As for the post study into which groups are randomized, (n=32 for each), i would expect the SD to be around 2.5, and the mean of the intervention group to be greater than control.

Hope this helps and any help is much appreciated.

 

[rogojel]

hi,
so you alteady have a sample size, what you can do is calculate the smallest significant difference in the means that your study can idenify - and then decide whether the study is worth doing. It will depend on the type of test you plan to do - so, is it a t-test, a paired t-test or some non-parametric test?

Regards