[FieldTrip] scaling frequency data for comparison

[Eelke Spaak]

Dear Jakob,

I can't give you any definitive answer on the power units of the
different freqanalysis algorithms, but you might find the function
ft_freqbaseline helpful. Absolute power values are usually not very
meaningful in neuroscience research, since they depend on a lot of
non-brain factors. Ft_freqbaseline performs a baseline correction
(either relative, absolute, or percentage) on your power values.
Across different spectral estimation methods, the relative power
change should remain the same (since it is unitless).

Best,
Eelke

2011/8/16 jakob kaiser <jakobala at hotmail.com>:
> Dear list,
> I wanted to compare the outputs from different ft_freqanalysis-methods to
> see if they make significant differences in the final test results of my
> data. This were the parameters used:
>
> cfg.output       = 'pow';
> cfg.channel      = 'all';
> cfg.method       = 'mtmconvol';
> cfg.taper        = 'hanning';
> cfg.foi          = 6:2:20;
> cfg.t_ftimwin    = ones(length(cfg.foi),1).*0.5;
> cfg.toi          = -0.5:0.05:2;
>
> WaveletConfig.method = 'wavelet';
> WaveletConfig.output = 'pow';
> WaveletConfig.foi = 6:2:20;
> WaveletConfig.toi = -0.5:0.05:2;
> (I chose these two methods because I found them used un papers relevant for
> my work)
> Now, as has been discussed on the list before, the scaling of the output
> data is very different. While the mtconvol-hanning-method gives me data
> points between roughly -2 an 1, with the wavelet-method I have data between
> roughly -800 and 600. From the (btw very helpful) tutorials and the previous
> list discussions I understand that this is a normal byproduct of
> the mathematical methods employed. However, for comparing different outputs
> it would be useful to scale both produced data sets to the same unit, so to
> speak. As both methods are supposed to measure the same thing, it should
> someone be possible to scale them to similar sizes, shouldn't it? Is there
> such a way? (Ultimately, I will have to compare the data results to results
> from a completely different program, which seem to have a third, different
> scaling altogether, so I am really puzzled how to handle the differences in
> scale)
> Generally, although this has been mentioned before, I am still not sure, if
> it is possible to put a certain unit to the results. Power is often reported
> in mV^2/Hz. If my initial data is in mV, would it be correct to say, that
> the hanning-method describe above delivers data in mV^2/Hz? Would this be
> correct for the Wavelet-method? Obviously, both cannot be true, because the
> scaling is extremely different. So, is it possible to specify a unit for any
> of these methods? Is there a way to convert the output
> of ft_freqanalysis-methods to this or another meaningful unit?
> I would be very grateful, if someone could give me a hint, how to interpret
> the data points here.
> Thank you for reading + thanks to the programmers for fieldtrip, which I
> like a lot
> Jakob
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>