Experimental design and analysis question
My name is Rowan and I've completed the most recent Homer 3 virtual training and have a couple of questions. I collected data using oxysoft and my experimental design is a series of cognitive tasks, with a baseline activity measurement. Previous members of my group have analysed the data by taking a mean oxy/deoxy value of each experimental condition subtracted from the mean values of the baseline measurement (done using oxysoft not Homer3). I essentially want to find out the relative changes in oxy/deoxy from each experimental condition, compared to a baseline. Some of the things I'm a little unclear on are as follows
1) I placed event markers during each measurement to signal the beginning and end of an experimental block. Some tasks have experimental blocks that differ in length. (as some tasks are self paced and others aren't). Given this can I still use the block average function to calculate HRF? (i.e. it seems to me the block average calculation relies on the blocks being of equal length) Some conditions also only consistent of one block (so would only have 1 block average to calculate the HRF rather than multiple blocks), is this a problem?
2) If this is possible would it make sense (analytically) to take the HRF function of each condition and subtract this from the HRF of the control condition, to isolate the activity created by the experimental manipulation
3) When I used Homer2 there was an option to export the raw concentration values (and I did use this in the beginning to perform the stats on). I can't see this function anymore, so I assume the HRF is what I would be interested in analysing?
4) Finally I am having the problem that when I use a high pass filter (0.01Hz), the activation becomes incredibly flat (i.e. the concentration differences become incredibly small). I remember from the lectures that this can happen, but I'm concerned that this is abnormally small. If you are not using drift correction in the GLM (i.e. using block averages), is there a way of correcting for drift, that would circumvent this problem (assuming it is a problem)
Sorry this is quite a long post, if would be more useful over email please let me know
1) you can certainly use the block average. Just make sure your tRange is long enough to capture the longest trial. But you shouldn't be placing stim marks at the end of a block, only at the beginning. Homer will see your stim marks at the end and assume that is the START of another stimulus. No problem with having a stim condition with only 1 trial. As for interpretation of averaging trials with different length stimuli, well, that is up to the researcher who designed the experiment. I don't have experience with variable length trials myself.
2) You are asking about subtracting the average of one condition from another. I understand that some people do do that. You can also just compare the averages of the two conditions and look for significant differences without subtracting the,=m. You should be able to find examples in the literature.
3) It is still possible to export HRFs in Homer3... I just don't recall how it is done and don't have Homer3 open right now. Hopefully someone else will answer this.
4) How long are you stimuli? A HPF of 0.01 Hz would normally not flatten out your HRF. But if your trial duration is 50 sec or longer then yes, this HPF would flatten out your HRF. It is hard to do drift correction on such long stimuli.
3) From Homer3 GUI: File -> Export -> HRF / Subject HRF Mean