Not clear from your message, but there are two likely issues. One is that repeating items may be confounded with other manipulations*. In that case the design is problematic. If not you can include "animation" as a fixed effect in the model. This may improve the senssitivity of your tests.
The effects will generalize only to those animations used in the study - if you wish to generalize beyond these you need to treat "animation" as a random effect being sampled. Thom * this can be avoided by counterbalancing or randomizing the occurance of the animations. Harold Hill wrote: > > I wish to us a repeated measures design to investigate the effects of > two types of exaggeration on the perception of animations. For both > types of animation I would like include a level that leaves the > orginal unchanged. Thus some trials would involve presentaion of > identical animations but be in different cells of the design. I guess > my default approach would be to ignore this fact, except perhaps to > plan to test that there is no difference between the equivalent cells. > However it seems you should be able to take advantage of the repetion > either 1) By not duplicating the trials but using the same data for > both cells of the design thus saving your observer's some trials. I > guess this invalidates the default model of sources of variance 2) > Include duplicate trials, but in some way use the fact they are > equivlaent to give you a better estimate of the variance associated > with the equivalent cells. Not sure how you would do this though. > I hope this is a vaguely interesting question for this group and I > would much appreciate any thoughts/references. I can't believe I am > in a unique situation as no treatment must be a fairly common level to > want in for a variety of variables but it is not something that I have > seen covered. Please note I am not a statistician but a humble > psychologist - please tailor any answers appropriately. This is what > also makes me baulk at trying to implement an explicit model of > sources of variances for this situation though I guess this is what > you may recommend. > Thanks, > Harold . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
