Averaging duplicate samples

[lu_king]

I have a data set that contains date of sample, site locations, main samples duplicate samples, and PH levels. I am trying to average the non-zero duplicate samples with the main samples, specific to the location (matching SA to SA, and AU to AU). Any advice would be appreciated!

For example,
Date location sample ph level
5/6 SA Main 2.5
5/7 SA Dup 6.467
5/6 AU Main .04
5/7 AU Dup 0

[Tom]

Should be simple.

proc summary data=have;
  by date location;
  var ph_level;
   output out=want mean=mean_ph_level;
run;

I don't understand this part "the non-zero duplicate samples with the main samples".  0 should be a perfectly valid value of pH.  Why would you want to exclude them?

[lu_king]

Hi Tom, thanks for your help! I didn’t want the zero averaging with the main sample because most of the time a zero is due to a user error/malfunction (the sampler is clogged, someone forgot to run it, etc). Usually a note is made but sometimes it isn’t, I figure if the duplicate yielded a zero then there was no need to average it with the main sample.

[Quentin · Accepted solution]

Maybe add a WHERE statement to Tom's code, e.g.

proc summary data=have;
  by date location;
  var ph_level;
  output out=want mean=mean_ph_level;
  where ph_level ne 0;
run;

or 

proc summary data=have;
  by date location;
  var ph_level;
  output out=want mean=mean_ph_level;
  where NOT (location='Dup' and ph_level=0) ;
run;