/******************************* TOSA_CD **********************************/
/***This program is used to calculate centroid based topological change***/
/***********************Author: Nan "Wendy" Li ***************************/
/***********************wendyli0611@gmail.com*****************************/
/****************************May 16, 2015*********************************/

proc printto log="C:\temp\Save_my_log_please.log" new; /***Creat a log file***/
%macro TOSA_CD(sample,data); /*Input data*/
data _null_; 
	 call symputx('ndat',count("&data",'|')+1);
run;

%do k=1 %to &ndat; /*Read data from an Excel spreadsheet*/
  %let dat=%scan(&data,&k,'|');
  data dis_&dat._mean; set _null_; run;

  proc import out      = &dat
  	          datafile = "&path.\&sample..xls" 
              dbms     = xls REPLACE;
              sheet    = "&dat"; 
              getnames = yes;
  run;

  proc contents data=&dat out=content;run; /*Read raw data into a SAS data table*/
  proc sort data=content; by VARNUM;run;
  proc sql;
       select name into: vlist separated by ' ' from content;
       select count(name) into: nv from content;
       select distinct(NOBS) into: num from content;
  run; quit;
  %let num=%trim(%left(&num)); /*Initialize parameters*/

  proc corr data=&dat cov out=cov; /*Calculate covirance matrix*/
       var &vlist;
  run;

  data cov; set cov; /*Save covirance matrix*/
       keep &vlist;
       if _n_ in (1:&nv);
  run;

  proc means data=&dat; output out=&dat._mean;run; /*Calculate the centroid of the raw data*/
  data &dat._mean; set &dat._mean;
       drop _TYPE_ _FREQ_ _STAT_;
       if _n_=4;
  run;

  proc iml;
       use cov;
       read all into cov;
       close cov;
       cov_inv=inv(cov); /*Calculate the inverse matrix of original covirance matrix*/

       use &dat;
       read all into g;
       close &dat;

       use &dat._mean;
       read all into m;
       close &dat._mean;
		
       %do i=1 %to # /*Calculate the mahalanobis distance between any data point and the centroid*/
           c1=g[&i,];
           c&i=c1-m;
           t_c&i=t(c&i);
	   _tmp&i=c&i*cov_inv*t_c&i;
           _z&i=sqrt(_tmp&i);
           create _z&i from _z&i[colname={"dis_&dat._mean"}];
           append from _z&i;
       %end;
 quit;

 %do i=1 %to # /*Calculate the results*/
     data dis_&dat._mean; set dis_&dat._mean _z&i;run;
 %end;
 proc datasets nolist; delete _z:; run; quit;
%end;

data test; merge dis_input_mean dis_output_mean; /*Output the results into a data table*/
   diff_sq=(dis_output_mean-dis_input_mean)**2;
run;
proc means data=test; /*Calculate the mean distance difference*/
     var dis_output_mean dis_input_mean diff_sq; 
     output out=means(drop=_TYPE_ _FREQ_);
run;
data means; set means;
  if _STAT_ in ('MEAN' 'N');
  if _STAT_='N' then _STAT_='num_of_s';
  if _STAT_='MEAN' then diff=sqrt(diff_sq); 
run;

proc export data= work.means /*Export the results*/
            outfile= "&path.\TOSACD_result.xls" 
            dbms=excel replace;
            sheet="TOSACD_result"; 
run;
%mend;

%TOSA_CD(sample=sample,
         data=input|output,
         path=C:\test);

proc printto; title; run;