reading values that include both numbers and letters from cvs using proc import

[yoavgn]

I need to read a cvs file that has value such as "1 female" (for the sex variable) and I would like to read this value into a numeric variable that will have the value of 1 for females

Thanks

Yoav

[Tom]

When reading from a delimited file you want to use list mode input.  With list mode the width of the informat is ignored.  So your choices are limited to reading it into a character variable or creating your own informat.

For the first option you could then convert the string to a number using the INPUT() function.  You could also define the character variable to have a length a just 1 and the other characters will be ignored when the line is read since there will be no place the store them.

data test1;
  infile 'myfile.csv' dsd trucover ;
  input id gender :$1. ;
  gender_code = input(gender,1.);
run;

proc format ;
  infile gender (upcase)
   '1 FEMALE'=1 '2 MALE'=2 
  ;
run;

data test2;
  inflie 'myfile.csv' dsd truncover ;
  input id gender_code :gender. ;
run;

[yoavgn]

Thanks a lot Tom

The problem is that I have many values like that. For example income is coded "1 21k-25k" "2 26k-30K" etc. Is there a way to extract just the numerical part of the value?

[Tom]

Read the values as the strings they are and then create new numeric variables from them.

data step1;
  infile 'myfile.csv' dsd truncover;
  length id cvar1-cvar10 $80 ;
  input id cvar1-cvar10;
run;
data step2;
  set step1;
  array old cvar1-cvar10;
  array new nvar1-nvar10;
  do index=1 to dim(old);
    new[index]=input(scan(old[index],1,' '),32.);
  end;
run;

[yoavgn]

Meanwhile I found a real easy solution. One way to to it is, for example when the value for female is "2 female" your state sexn=input(sex,1.) and if turns sex into a mumeric variable sexn.