/********************************** TOSA_CV ************************************/
/*This program is used to calculate centralized vector 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_CV(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 vector_&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; /*Calculate the inverse matrix of original covirance matrix*/
       use cov;
       read all into cov;
       close cov;
       cov_inv=inv(cov);

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

       use &dat._mean;
       read all into m;
       close &dat._mean;

       %do i=1 %to # /*Calculate the angle between any two vectors starting from the centroid*/
           c1=g[&i,];
           t_c1=t(c1);
	   t_m=t(m);
	   _tmp=c1*cov_inv*t_m;
	   _r1=c1*cov_inv*t_c1;
           _r2=sqrt(_r1);
	   _q1=m*cov_inv*t_m;
	   _q2=sqrt(_q1);
	   _rq=_r2*_q2;
	   _rq_inv=inv(_rq);
           _z&i=_tmp*_rq_inv;
           create _z&i from _z&i[colname={"vector_&dat._mean"}];
           append from _z&i;
      %end;
quit;

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

data test; merge vector_input_mean vector_output_mean;
   diff_sq=(vector_output_mean-vector_input_mean)**2; /*Calculate the suqare sum of the angle difference between input and output data*/
run;
proc means data=test; 
     var vector_output_mean vector_input_mean diff_sq; 
     output out=means(drop=_TYPE_ _FREQ_);
run;
data means; set means; /*Calculate the mean difference*/
  if _STAT_ in ('MEAN' 'N');
  if _STAT_='N' then _STAT_='num_of_s';
  if _STAT_='MEAN' then diff=sqrt(diff_sq); /*Calculate the result*/
run;

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

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