*   read a dbf file;
*   pobf.sas;
*   data source is the dma filesat http://oig.hhs.gov/fraud/exclusions/database.html;
filename file1 '/temp/po.dbf';
libname db3 '\temp\po';
options compress=yes linesize=72;
proc dbf db4=file1 out=db3.po;
run;
options obs=max;
data sel;
   set db3.po;
   length kflag 4;
   length lnam2 $2;
   length lnam1 $1;
   length lnam3 $3;
   length vendno $6;
   
   vendno = compress(vendor_n);
   lnam2= substr(vendno,1,2);
   lnam1 = substr(vendno,1,1);
   lnam3= substr(vendno,1,3);
   kflag = 0;
   if lnam2 = "E1" then kflag = 1;
   if lnam2 = "E2" then kflag = 1;
   if lnam2 = "E3" then kflag = 1;
   if lnam2 = "G2" then kflag = 1;
   if lnam1 >= "0"  and lnam1 <= "9" then kflag = 1;
   if lnam3 = " E1" then kflag = 1;
   if lnam3 = " E2" then kflag = 1;
   if lnam3 = " E3" then kflag = 1;
   if lnam3 = " G2" then kflag = 1;
   if lnam2 >= " 0"  and lnam2 <= " 9" then kflag = 1;
   if kflag = 0 then delete;
run;
proc contents data=sel;
proc summary data=sel;
      var invamoun;
      output out=s n=n sum=tot min=min max=max;
run;
proc print data=s;
      title1 'extract statistics';
      run;
proc sort data=sel;
      by vendno;
proc summary data=sel;
      by vendno;
      var invamoun;
       output out=s2 n=n sum=tot min=min max=max;
data s2a;
       set s2;
       if tot < 1000 then delete;
        
proc print data=s2a;
      title1 'summary statistics by vendor';
      title2 'only vendors with more than $1,000';
      run; 
options obs=max;
data test;
      set sel;
      length ct 4;
      length amt $9;
      length digit $1;
      length tdigit $1;
      amt = put(invamoun,9.2);
      digit = ' ';
      tdigit = substr(amt,1,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,2,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,3,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,4,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,5,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,6,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,7,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,8,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      tdigit = substr(amt,9,1);
      if digit = ' ' and tdigit not = ' ' and tdigit not = '0' and tdigit not = '.' then digit = tdigit;
      ct = 1;   
proc sort data=test;
      by vendno digit;
proc summary data=test;
      by vendno digit;
      var ct;
      output out=s3 n=n sum=tot;
proc summary data=test;
      by vendno;
      var ct;
      output out=s4 n=n4 sum=tot4;
data both;
      merge s3(in=s) s4(in=t);
      by vendno;
      if s and t;
      if n4 < 10 then delete;

data both2;
       set both;
       length expect 4;
       keep vendno digit n n4 expect;
       if digit = '1' then expect = int(log(1+(1/1))* n4);
       if digit = '2' then expect = int(log(1+(1/2))* n4);
       if digit = '3' then expect = int(log(1+(1/3))* n4);
       if digit = '4' then expect = int(log(1+(1/4))* n4);
       if digit = '5' then expect = int(log(1+(1/5))* n4);
       if digit = '6' then expect = int(log(1+(1/6))* n4);
       if digit = '7' then expect = int(log(1+(1/7))* n4);
       if digit = '8' then expect = int(log(1+(1/8))* n4);
       if digit = '9' then expect = int(log(1+(1/9))* n4);
       
proc print data=both2;
      title1 'leading digit summary by vendor';
      

run;