How to Select Study Participants to Include in Your Analyses

The day-level metrics that you created in previous steps will now be averaged over multiple days to produce one record per person. Only data from valid persons will be summarized. A valid person is a participant who has a sufficient number of valid days. Criteria for a valid day and a valid person can be defined by the analyst. In the sample code, a valid day is a day in which the participant wore the monitor for at least 10 hours and a valid person is a participant who has at least 4 valid days. Use the PROC SUMMARY statement to count the total number of valid days and the DATA step to identify valid persons.

Sample Code

*-------------------------------------------------------------------*;
* Define a valid day and a valid person.                            *;
* NOTE:                                                             *;
* to change the definitions on the number of wear hours (wear_hr)   *;
* required for a valid day, or the number of valid days (valdays)   *;
* required for a valid person, please modify the statements below.  *;
*-------------------------------------------------------------------*;

data pam_day;
  set pam_perday;
  valid_day=(wear_hr>=10 );  * assign valid day hours criterion here *;
  format valid_day yesno. ;
  label valid_day= '10+ hours of wear (yes/no)' ;
run ;

proc summary data=pam_day;
  by seqn;
  var seqn;
  where valid_day= 1 ;
  output out =valid
         n=valdays;     * number of days with 10+ hours of wear *;
run ;

data pam_day;
  merge pam_day(in=inall) valid;
  by seqn;
  if inall;

  if valdays= . then valdays= 0 ;
  label valdays= 'Number of days with 10+ hours of wear' ;

  valid_person=(valdays>=4 );  * assign valid person days criterion here *;
  format valid_person yesno. ;
  label valid_person = 'At least 4 days with 10+ hours of wear (yes/no)' ;

  drop _freq_ _type_;
run ;