when should I take the mean in my analysis

Hello,

I am currently analysing a vision experiment and have same problems in
linking the stuff I read in the books with the experiment.

The data I have:

- 10 subjects, each subject sees 144 images and during watching an
image we record something.

What we have done so far:

- following the literature we take for each subject and in each image
the mean over the things we record (recognition times, saccade length
etc.) - results in 10 * 144 data points

The purpose of the study:

The 144 images can be devided in 6 categories containing 24 images- the
categories arise due to accounting for 2 factors (2D/3D image, three
different kind of images - 2*3=6). We want to find out whether there is
a difference between the categories. Also important to know, there are
only 24 underlying images and due to transformation we get 6 different
stimuli for each image. So, there is a dependence between categories.
Each image is by modification shown in 6 categories.

I would suggest to do a Friedman test (or a ANOVA) with two factors
and use a dependent analysis, because then we can account for the
variance between subject. That would mean that I take the mean over the
images in each category, get one number per category and therfore 6
numbers per subject - then I could perform a Friedman test.

But what about the dependence between images. When I take the mean
over images, I will loose that, although it is inside the data. But I
do not know how I can include it in such an analysis.

As nobody at the University can help me, I hope to find some inspiring
ideas in the newsgroup.

If something is unclear, then please ask. It was quite hard to describe
the design shortly.

Thank you in advance
lisra

.

Loading