matlab - Feature Extraction through PCA -

i'm trying select subset of features data contains 2000 of them 63 samples. know how pca in matlab. used 'pcacov' , returns eigenvectors , eigenvalues too. however, don't know how select features want. mean if features aren't labeled, how can select features ? or returned in same order ?

how can select features ?

if call like

[pc,variances,explained] = pcacov(covx) 

then principal components vectors in first return argument variances in second return argument. in correspondence , sorted significant least significant.

or returned in same order ?

you can assume if function says so, otherwise it's not safe assume , can like.

[varsorted,varsortedinds] = sort(variances,'descend'); pcsorted = pc(:,varsortedinds); 

and varsorted , pcsorted in order least significant.