| T O P I C R E V I E W |
| ch186 |
Posted - May 07 2014 : 09:59:57 AM
I am using eventstats and then matlab (ROI_timecourse.m that runs BIAC roi function, roianalysis_timecourse_iris) to do an ROI analysis of the amygdala. I want to weight early trials across my conditioning runs and across my generalization runs with an exponential decay function. In our experiment, amygdala activation usually shows up only during the first few trials of conditioning, and again during the first run of generalization testing.
Where should I do this weighting? Is it something that I would do in the 3-column files before even running eventstats, or can it be done later on when running the actual ROI script?
Thanks! |
| 4 L A T E S T R E P L I E S (Newest First) |
| syam.gadde |
Posted - May 07 2014 : 10:45:13 AM
The condition_n.bxh files are just the "N" -- i.e. the number of trials that went into each voxel used in the analysis. Can be used voxel-by-voxel to calculate other statistics, but that and all the other outputs of eventstats (with the exception of condition_trials.bxh) are still collapsed by condition.
|
| ch186 |
Posted - May 07 2014 : 10:40:50 AM
You're correct, no 3 column files are used. I mixed that up with the regular analysis with fsl feat. I do use the condition_avg.bxh from eventstats for the ROI scripts so I guess that info is lost. In the output of eventstats, are the files labeled "condition_n.bxh" the concatenated 4-D files for those individual trials? If so, could I just weight those somehow and re-calculate the avg.bxh? Then I wouldn't need to run all of eventstats over again. |
| ch186 |
Posted - May 07 2014 : 10:34:07 AM
You're correct, no 3 column files are used. I mixed that up with the regular analysis with fsl feat. I do use the condition_avg.bxh from eventstats for the ROI scripts so I guess that info is lost. In the output of eventstats, are the files labeled "condition_n.bxh" the concatenated 4-D files for those individual trials? If so, could I just weight those somehow and re-calculate the avg.bxh? Then I wouldn't need to run all of eventstats over again. |
| syam.gadde |
Posted - May 07 2014 : 10:26:30 AM
Just curious, where do the 3-column files come into play? Are they used in a GLM? As far as I remember, eventstats and the ROI tools don't use them, so just making sure there isn't more to the analysis than those.
Does the ROI script take a concatenated set of trials or are the trials averaged already? If they are averaged already by condition, you've lost the ability to weight the individual trials, so you have to do it before then.
eventstats has an option:
--extracttrials
If this option is specified, the program will write out epochs for
*all* extracted trials to a file PREFIX_QUERY_trials.bxh. This file
will be a 5-D image file where the 4th dimension goes across time
points within an epoch, and the 5th dimension represents the global
trial number.
This option hasn't been used widely, but could give you the ability to weight the trials. You'd probably have to calculate the final statistics yourself.
|
|
|
