shift.m

function [clean_data,shift_data,pp_test]=shift(index,train_trial,Xplan)
    % p_test is our testing set
    K=4;D=97;L=length(index);inc=15;rate=floor(150/15);N=rate*L;

    train_hold=zeros(D,200,N,K);
    clean_hold=zeros(D,200,L,K);
    diagonal=1:2:2*K;
    for k=1:K
        for l=1:L
            clean_hold(:,:,l,k)=train_trial(index(l),diagonal(k)).spikes(:,351:550);
            for r=1:rate
                train_hold(:,:,l*r,k)=train_trial(index(l),diagonal(k)).spikes(:,(351+(inc*r)):(550+(inc*r)));
            end
        end
    end

    % Baseline no data augmentation dataset
    clean_count=squeeze(sum(clean_hold,2));
    clean_combined=reshape(clean_count,D,K*L)';

    % With data augmentation
    spike_count=squeeze(sum(train_hold,2));
    combined_count=reshape(spike_count,D,K*N)';

    % labels
    train_labels=ones(K,N).*[1:K]';
    train_labels=reshape(train_labels',N*K,1);

    test_labels=ones(4,91).*[1:4]';
    test_labels=reshape(test_labels',4*91,1);

    clean_labels=ones(K,L).*[1:K]';
    clean_labels=reshape(clean_labels',L*K,1);


    rng(0)
    [coeff,score,pcvar,mu,v,S] = ppca(combined_count,8);
    S
    clean_coeff = ppca(clean_combined,8);

    clean_data=clean_combined*clean_coeff;
    clean_data(:,end+1)=clean_labels;
    clean_data=clean_data(randperm(K*L),:);

    shift_data=combined_count*coeff;
    shift_data(:,end+1)=train_labels;
    shift_data=shift_data(randperm(K*N),:);

    a=1:91;b=[a ,a+2*91,a+4*91,a+6*91];
    pp_test=Xplan*coeff;
    pp_test=pp_test(b,:);
    pp_test(:,end+1)=test_labels;


end