About Food-Tracking Task

I used function tfrscalo to make scalogram. (your link on the homepage didn’t work, so I found it using google I’m not sure it’s same as yours…) for all time, i execute tfrscalo and
***********************************
window=double(Data(channel,(sec-1.1)*1000+1:round(sec*1000)));
% 1100 sample = 1.1s 1ch ECoG data 64*1100

[TFR,F,WT]=tfrscalo(window',1:1100,7,10/Fs,120/Fs,10,0);
% scalogram for 1.1s data

for freqbin=1:10
for timebin=1:10
NORMTFR(freqbin,timebin)=mean(TFR(freqbin,timebin*100+1:(timebin+1)*100));
%down sampled to 10*10
end
end

normersp=zscore(NORMTFR,0,2);
% normalized scalogram matrix each freq.

erspfeat(time,channel,:)=normersp(:);
% 10*10 matrix to 1*100 matrix

erspfeature(tiem,:)=erspfeat(time,:);
% for all channel training data * 6400
*******************************************

I also normalized motion data. ( I used only right wrist, because you tracked right hand)
**********************
normmotion=zscore(RWRI);
% normalized motion data
*******************

And used scalogram,. With components of 50.
**********************
comp=50;
[xl,yl,xs,ys,traincoef,per,mse]=plsregress(erspfeature(1:training,:),normmotion(1:training,:),comp,'cv',10);
trainresult = [ones(validation,1)
erspfeature(training+1:training+validation,:)]*traincoef;
***********************

I followed the exactly same steps you described.
Re-refered -> tfrscalo -> downsample to 10*10 matrix -> make regression model except downsampled ECoG data.
I didn’t donwsampled ECoG data 1KHz to 250Hz, I used whole data..
But it didn’t work at all..

Q. Is there any step I missed or you didn’t mentioned ?

Q. Did you apply some filter or some pre-processing before you made the scalegram ? I directly used the data from neurotycho data(raw data). ( I didn’t care about chewing effect…)
Q. Could you let me know which day’s data did you used ?

Re: About Food-Tracking Task

Q. Is there any step I missed or you didn’t mentioned ?
Ans. We would like you to carefully read our paper and answers from other questioners at Neurotycho wiki http://wiki.neurotycho.org/Epidural-ECoG_Food-Tracking_Task. There is everything what we actually did in our study. Other questioners were able to resolve their wrong code with these answers. Different time sampling rate between motion data (120Hz) and eECoG signals (1kHz) might be due to your issues. In addition, we used the optimal number of PLS components, which are determined by the minimal predictive error sum of squares. Therefore, the number of PLS components is different in each experiment.

Q. Did you apply some filter or some pre-processing before you made the scalegram ? I directly used the data from neurotycho data(raw data). ( I didn’t care about chewing effect…)
Ans. As shown in our paper (2.3. Decoding paradigm), 1kHz of eECoG signals were band-pass filtered from 0.3 to 500Hz.

Q. Could you let me know which day’s data did you used ?
Ans. All data set uploaded at Neurotycho were used in our paper.