Back to the main page.

Bug 1379 - non-desired behaviour for lcmv with trial and filter input to ft_sourceanalysis

Reported 2012-03-16 15:33:00 +0100
Modified 2012-05-16 14:19:02 +0200
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Windows
Importance: P3 normal
Assigned to: Johanna
Depends on:
See also:

Johanna - 2012-03-16 15:33:12 +0100

Maybe this will be fixed with a new ft_sourceanalysis, but it was recently pointed out to me by Simon Hanslmayr that .pow is based on the ERP and not all the trials, even if the timelock structure you enter has trials as well as avg. The only way around this is to set cfg.rawtrial='yes' (a simple thing for the user to do) but then you get all the trials saved out the in .pow as well (which should be controlled by cfg.keeptrials I would think). So, even if the filter was correctly computed based on the average of the covariances (rather covariance of the average) by calling ft_sourceanalysis a first time with keepfilter='yes', then the second call to ft_sourceanalysis to apply this filter doesn't give the 'expected' (I would assume default) output. Suggestion: if cfg.grid.filter exists, and if timelock.trial exists, then force cfg.rawtrial to equal 'yes' by default.

Johanna - 2012-05-16 14:19:02 +0200

for the meantime (whilst waiting on more major changes to ft_sourceanalysis), update the documentation of this behaviour.