[FieldTrip] which baseline correction to choose for frequency analysis

[Stenner, Max-Philipp]

Hi Katrin,

it sounds like you will end up using 'depsamplesT' as your statistic (because you compare repeated measures = 2 consecutive time intervals from the same subject or trial) - why do you need to use baseline-correction then? The pairing should already account for differences in baseline power between subjects (or trials). If you do need a baseline correction, I would not choose such a short baseline in the frequency domain, but go for an average over a much longer (.5 to 1 second) interval, particularly if you intend to baseline-correct single trials (because a short baseline is more likely to project noise from the baseline into your interval of interest). Also, keep in mind that ft_freqanalysis computes power across a time window (specified by you in cfg.t_ftimwin), so that power between -.1 and 0 may not only reflect "baseline" power.

Best

Max



________________________________

Von: fieldtrip-bounces at science.ru.nl [fieldtrip-bounces at science.ru.nl]" im Auftrag von "KatrinH Heimann [katrinheimann at gmail.com]
Gesendet: Donnerstag, 29. Mai 2014 11:55
An: FieldTrip discussion list
Betreff: [FieldTrip] which baseline correction to choose for frequency analysis

Hey together,
another rather conceptual question:
I am wondering about the baselinecorrection for the frequencyanalysis.
I want to compare two timeperiods of one second each, lying right after each other.
In the preprocessing I have baseline corrected them using 100 ms before the first piece (last part of a fixationcross period).
Do I use the same part for the frequencyanalysis? Will say, do I run the analysis let's say from -0.5 (0 point is end of fixation cross) to 2.5 s, correcting with [-0.1 0] and then extract the two different conditions by statistical extraction (each 1 second long) ? Or do I run two analysis on (a bit more than) one second each, using the whole trial as baseline?

thanks a lot for your advice again! Best
Katrin