Back to the main page.

Bug 343 - purpose of baseline in ft_sourceanalysis?

Reported 2010-12-24 13:59:00 +0100
Modified 2011-05-11 14:48:24 +0200
Product: FieldTrip
Component: inverse
Version: unspecified
Hardware: All
Operating System: All
Importance: P1 normal
Assigned to: Johanna
Depends on:
See also:

Johanna - 2010-12-24 13:59:30 +0100

Jan-Mathijs asked me in relation to bug 199: if all I want is dip.pow from beamformer_lcmv, why not just use inv(lf*InvCy*lf) outside of calling ft_sourceanalysis? My reason is that the filter I compute is not usually based on the same Cy as that to which I would apply the filter. I prefer to compute the filter from an average of two covariances (e.g. active and baseline, or condition1 and condition2), then apply this filter to each covariance separately, in order to feel justified in comparing the conditions in source space. Yes, I could easily make this average happen in my own external code, after calling ft_timelockanalysis and before calling ft_sourceanalysis, but I mention this since a framework for entering a 'baseline' exists in ft_sourceanalysis and initially confused me. Currently, 'baseline' is only entered to ft_sourceanalysis a) for use with randomization or permutation, _and_ b) as multi-trial. I think it might be a useful option to allow for baseline.cov to be averaged with data.cov for computing Cy (either with Ntrials>1 or =1). For memory reasons like with data.cov, I prefer baseline.cov to be from 'keeptrials=no', so allow for that type of input? I have already hacked my copy of ft_sourceanalysis to do this, and am happy to share if this is reasonable; otherwise I can do these steps in my own external code if this should not be a supported option.

Johanna - 2011-05-04 15:58:25 +0200

the solution to bug 637 will solve this question of how to combine data using FT functions _after_ call to timelockanalysis or freqanalysis, for use to input to ft_sourceanalysis. this 'worksforme'.

Johanna - 2011-05-11 14:47:22 +0200

further information: use of cfg.offset with ft_redefinetrial will also be used to allow ft_appendtimelock to work over trials of data with time-windows of different times but equal lengths.