Multiple observations...

tomm

New Member
#1
Hi,

I have a repeated measures design with two within subject variables. Under each of the four(2x2) conditions, I measure completion time for a simple interaction task. Since the task only takes a few seconds, I take about 80 measurements for each condition.

I want to run a couple of a priori comparisons between conditions. If I understand things correctly, I can't run these tests on all of my data; instead I need a single value for each subject under each condition. (I guess because the individual measurements are not independent?)

It seems to me that by only doing the comparison on the means of all my observations, I'm throwing away a lot of information. Is this really the case? Does it not matter how good my individual estimates are? Is there any way I can make use of this data?

Thanks.

Tom.
 

JohnM

TS Contributor
#2
The reason why you can't do the comparison of conditions is that the measurements are not independent of each other and therefore the estimate of error variance within each condition is too low.