| MATCORR | Find matching rows in two matrices and their corrs. Uses the Hungarian (default), VAM, or maxcorr assignment methods. (Follow with matperm to permute and sign x -> y). |
| Usage: | >> [corr,indx,indy,corrs] = matcorr(x,y,rmmean,method,weighting); |
| Inputs: | |
x |
first input matrix |
y |
matrix with same number of columns as x |
| Optional inputs: | |
rmmean |
When present and non-zero, remove row means prior to correlation {default: 0} |
method |
Method used to find assignments. 0= Hungarian Method - maximize sum of abs corrs {default: 0} 1= Vogel's Assignment Method -find pairs in order of max contrast 2= Max Abs Corr Method - find pairs in order of max abs corr Note that the methods 0 and 1 require matrices to be square. |
weighting |
An optional weighting matrix size(weighting) = size(corrs) that weights the corrs matrix before pair assignment {def: 0/[]->ones()} |
| Outputs: | |
corr |
a column vector of correlation coefficients between best-correlating rows of matrice x and y |
indx |
a column vector containing the index of the maximum abs-correlated x row in descending order of abs corr (no duplications) |
indy |
a column vector containing the index of the maximum abs-correlated row of y in descending order of abs corr (no duplications) |
corrs |
an optional square matrix of row-correlation coefficients between matrices x and y |
| Note: | outputs are sorted by abs(corr) |
| Authors: | Scott Makeig & Sigurd Enghoff, SCCN/INC/UCSD, La Jolla, 11-30-96 |