%==================================================
%
% LLSimpute - Local Least Squares Imputation 
%
% (Missing Value Estimation Package)
%
% Author: Hyunsoo Kim
% Date: Fall/2003 - Spring/2004
% E-mail: hskim@cs.umn.edu
% Personal homepage: 
%     http://www.cs.umn.edu/~hskim
% Reference: Missing value estimation for DNA microarray gene 
%     expression data: Local Least Squares Imputation, H. Kim, 
%     G. H. Golub, and H. Park, Bioinformatics, to appear, 2004.
% This software may be free downloaded from site:
%     http://www.cs.umn.edu/~hskim/tools.html
% License:
%     It is free for academic or nonprofit insistutions. 
%     All right is reserved regarding commecial usage. 
%     Please consult if you try to use this package for 
%     commercial purpose.
% Comments:
%     Please let me know if you have done any improvement. 
% 
% Sample Usage:
%     E=impute_rowavg(0,miss_matrix,10); 
%
% Description:
%
% function E=impute_rowavg(set,miss_matrix,minexp) 
%
% Input parameter: 
%     set - the number of set,  
%             If the miss_matrix came from miss0.mat, then assign set=0
%     miss_matrix - a matrix that has missing values (1e99)
%     minexp - if the number of non-missing values in a gene (row) is less than
%             minexp, then report the gene (row).
% Output parameter:
%     E - the estimated matrix 
%
%====================================================
function [E] = impute_rowavg(set,miss_matrix,minexp)
global m n big;

%-----------------------------------------
% get input matrix E that contains row average
%-----------------------------------------
fprintf('Generating row-averaged E...\n');

E=miss_matrix; gene0=[]; gene1=[];
for i=1:m
      missidxj=find(miss_matrix(i,:) == big);
      if (length(missidxj) > 0)
         nomissidxj=find(miss_matrix(i,:) < big);
         exp=length(nomissidxj);
         if(exp==0) 
             %fprintf('%d ', i); 
             gene0=[gene0 i];
         else
             % row average
            avg=mean(miss_matrix(i,nomissidxj)); 
            E(i,missidxj)=avg;
            if(exp < minexp)
                gene1=[gene1 i];
            end
        end   
     end
end
fprintf('\n');

% save bad genes
fnout=sprintf('gene0_%d.csv', set);
csvwrite(fnout, gene0);
fprintf('the number of genes that have no non-missing entries: %d\n',length(gene0));
fnout=sprintf('gene1_%d.csv', set);
csvwrite(fnout, gene1);
fprintf('the number of genes that have less than %.2f non-missing entries: %d\n',minexp,length(gene1));

return;