Please please can you help me...

[whitelilly]

Claire Robinson 30 May at 16:07
My heid is about to spontaneoulsy combust and I am so stuck:

If you have 2 mins, can you have a wee read please?

I am doing a pretty basic within-participants experimental design. I am looking to see whether scores on cognitive performance differ significantly when participants are given an easy task first and then given a hard task. As you know, cognitive tests measure diffierent brain functions so I am going to give two seperate cognitive tasks (each has parallel forms) to show the effects of any sig result are not due to the cog test.

So, a particpant might get this:

easy task > cognitive test 1a > difficult task > cognitive test 1b > easy task> cognitive test 2a > difficult task > cognitive test 2b

when I am doing my analysis, will I have to do two seperate t-tests, one for Cog test 1 (change score between 1a and 1b) and one for Cog test 2 (change score between 2a and 2b)...

Also, will this affect my power calc? I am useless, I know!

[Rob]

Well, if you want to do some sort of comparison you might want to investigate a paired sample t-test?

[nostawnevets]

It would appear that you would yes. Rather than worry too much about your power calculation you can just correct for using more than one test using the Bonferroni method, which is dead easy. Give it a quick google. Alternatively use an ANOVA instead of t-tests, but that would affect your power calculation as ANOVA's need more subjects to get significant results. generally though, using as few tests as possible is to be advised.