While reading from a file, choose specific fields and compare them with another file in Unix bash?

Assuming you want to compare lines in file b to corresponding lines in file a (i.e. that both files have the same number of lines and you're comparing line-by-line), you can use a Perl script to do this for you:

#!/usr/bin/perl
$file_a = "/path/to/file_a";
$file_b = "/path/to/file_b";
open $fa,'<',$file_a or die "Failed to open file $file_a: $!\n";
open $fb,'<',$file_b or die "Failed to open file $file_b: $!\n";
@file_a = <$fa>;
@file_b = <$fb>;
close $fa;
close $fb;
for (0..$#file_a){
    ($col_c,$col_d) = (split / /,$file_a[$_])[2,3];
    $col_z = $file_b[$_];
    $to_display = "$col_c $col_d";
    if($col_z eq $col_d){
        $to_display .= " $col_z";
    }
    else{
        $to_display .= " pin";
        if ($col_z gt $col_c and $col_z gt $col_d ){
            $to_display .= " lon";
        }
        elsif($col_z lt $col_c and $col_z lt $col_d) {
            $to_display .= " sma";
        }
    }
    print "$to_display\n";
}

Save the above file as /path/to/myscript (after modifying the locations of file_a and file_b to your true locations), then make it executable: chmod +x /path/to/myscript and finally invoke it: $ /path/to/myscript.

As I said in my comment, I am not entirely sure what you are trying to do, you seem to want to always print fields c and d and to compare only d with z. If so, the solution below should work.

$ paste a b | awk '{print $3,$4,$6}' | head -n 1; paste a b | tail -n +2 |
   while read a b c d e z; do 
    echo -n "$c $d"; 
    if [ "$d" -lt "$z" ]; then 
      echo "pin sma"; 
    elif [ "$d" -gt "$z" ]; then 
      echo "pin lon"; 
    else echo $z; 
    fi; 
   done 

Run on the example files you have provided, this gives:

c d z
2 33
6 5pin lon
0 9pin lon

Explanation

Since you want an arithmetic comparison, the headers will break the script. However, I imagine you would like to have them in the final output. So, first I print them:

paste a b | awk '{print $3,$4,$6}' | head -n 1;

So, now we want to process the rest of the fields, skipping the header. So, we paste the files again, skip the headers (tail -n +2) and process the fields.