Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 1 down vote favorite

I tested reactions to pictures under five conditions (neutral, high status, low status, high adhesion, low adhesion), using an EEG of 16 channels. The pictures were presented in random order for 15 times under each condition. So at best I have data from 75 trials for each subject in each channel. But as some electrodes sometimes made me some problems, I can't use the data from all of the EEG channels for all of the subjects. And as an EEG is sensitive to movements of the head, especially to eye movements, I furthermore had to reject some of the trials in all of the EEG channels for some of the subjects. So, on the whole, I neither have the same number of channels for each subject nor the same number of trials in the lasting channels for all of the subjects and therefore an unbalanced design.

I split-up the data in 12 time segments per EEG channel and condition and I'm interested in the divergent effects of the conditions on the EEG waves in each of these time segments. So, I want to compare the data (trials) from all the subjects for condition "status high" in channel 1 for time segment 1 with the data from all these subjects for condition "status low" in channel 1 for time segment 1. And the data from all the subjects for conditions "status high" in channel 1 for time segment 2 with the data from all these subjects for condition "status low" in channel 1 for time segment 2. And so on, until the pairwise comparison for time segment 12 in that first channel. And then, I want to do the same with channel 2 and channel 3 and so on, until channel 16. And then, I do the same with "adhesion high" vs. "adhesion low" for all of the channels.

Could I run these tests with the help of a Linear Mixed Model? What would be the fixed and random effects? Or would a multivariate approach be the better choice?

I would be so grateful, if you could sacrifice a little bit of your time for my problem. Please excuse my maladroit wording.

share|improve this question
Could you clarify what the response variable is for one channel on one subject in one trial? Are there multiple measurements, and if so, what is the relationship between an individual measurement and a 'time segment'? – atiretoo Jul 2 '12 at 22:52
Hey atiretoo! the response variable is a long line of points of different height for each subject in each channel. these lines of points are recorded at the same time in all of the 16 eeg-channels. so, an individual measurement consists of long lines of points of different height in 16 eeg-channels at the same time. This line of points can then be split into the same 75 trials per channel. And of these 75 trials 15 belong to the condition "status high", 15 to the condition "status low" and so on... – Hannah Jul 4 '12 at 21:23
in fact, I'm interested in the mean height of the line in each condition (over all of the trials for this condition from all of subjects) and I want to compare these means, but I have to differentiate between the 16 eeg-channels and the 12 time segments per trial in doing so. – Hannah Jul 4 '12 at 21:23
OK, the response of any one channel would be a linear mixed model. I'm not sure what to do about the 16 channels - the methods I know don't handle multivariate responses. Other's might have better ideas. – atiretoo Jul 7 '12 at 15:00

Know someone who can answer? Share a link to this question via email, Google+, Twitter, or Facebook.

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.