Macro to [Helmert transform](http://en.wikipedia.org/wiki/Helmert_transformation) (forward mean-difference) panel data

HELMERT.sas

/*******************READ ME*********************************************
* - Macro to Helmert transform (forward mean-difference) panel data -
*
* SAS VERSION:    9.4.0
* DATE:           2013-05-05
* AUTHOR:         eddyhu at the gmails
*
****************END OF READ ME******************************************/

%macro HELMERT(dsetin = &syslast., 
               dsetout = ,
               byvar = none,
               vars = ,
               lag = yes
               );

/*****************************************************************
*  MACRO:      HELMERT()
*  GOAL:       Macro to Helmert transform (forward mean-difference) panel data
*  PARAMETERS: libname     = SAS library (default USER)
*              dsetin      = input dataset containing variables to helmert transform
*              dsetout     = output dataset (leave blank to overwrite DSETIN)
*              byvar       = variable(s) used to form groups (leave blank for total sample)
*              vars        = variable(s) that will be helmert transformed
*              pass        = Postgre user password
*              db          = Postgre database
*              format      = DATA step format statement (optional)
*              rename      = DATA step rename statement (optional)
*              debug       = if y then send BULKLOAD trace info to SAS log
*                            useful for figuring out if you have badly formatted
*                            columns, and if so what specific row is being rejected
*
* DESCRIPTION:     This macro takes the input variables and performs the Helmert transform.  
*                  The Helmert transform is a forward mean-differencing transformation, in   
*                  this case normalized and computed based on the BYVAR macro variables      
*****************************************************************/

    %if &dsetout = %then %let dsetout = &dsetin;
    
    %let xn=1;
    %let h_vars = ;
    %let x_vars = ;
    %let z_vars = ;

    /* Forward sum */
    proc expand data=&dsetin out=xtemp;
        by &byvar;
        %do %until (%scan(&vars,&xn)= );
            %let token = %scan(&vars,&xn);
            %let h_vars = &h_vars h_&token;
            convert &token = h_&token / transform = (reverse sum reverse lead 1);
            %let xn = %EVAL(&xn + 1);
        %end;
        
    /* Count number of elements in sum */
    proc expand data=xtemp out=xtemp2;
        by &byvar;
        convert TIME = N / transform = (reverse);
    
    %let xn=1;
    /* Forward mean difference and weighting */
    data &dsetout;
        retain &byvar &vars &h_vars;
        set xtemp2;
        by permno;
        %do %until (%scan(&vars,&xn)= );
            %let token = %scan(&vars,&xn);
            h_&token = (&token-h_&token/N)*(N/(N+1))**.5;
            %let xn = %EVAL(&xn + 1);
        %end;
        drop TIME N;
    run;
    
    %if &lag = Yes | &lag = yes | &lag = Y | &lag = y %then %do;

        %let xn=1;
        /* Lags */
        proc expand data=&dsetout out=&dsetout;
            by &byvar;
            %do %until (%scan(&vars,&xn)= );
                %let token = %scan(&vars,&xn);
                %let x_vars = &x_vars x_&token;
                %let z_vars = &z_vars z_&token;
                    convert h_&token = x_&token / transform = (lag 1);
                    convert &token = z_&token / transform = (lag 1);
                %let xn = %EVAL(&xn + 1);
            %end;
        
        data &dsetout;
            retain &byvar &vars &h_vars &x_vars &z_vars;
            set &dsetout;
            drop TIME;
        run;

    %end;
            
    /*Delete temporary datasets created during macro execution*/
    proc datasets library=work nolist;
        delete xtemp xtemp2; quit; run;

%mend;