Hello,   I'm working with data that includes up to 5 event date & time variables per row. Missing data is an issue. The 5 events should have occurred in chronological order, (e.g., Event1 before Event2 before Event3 . . .). For each event, there may be a SAS date variable only, a SAS time variable only, both variables or missing data.   I need to flag observations if the events are not in chronological order.  For observations with complete information for each event, this is not a problem.  I can't figure out how to work with rows that have missing data. I would like to generate 2 variables: (1) flag if there is good chronological order for the available date & time variables and (b) flag for good chronological order for the available SAS dates only.   Any suggestions will be most appreciated.   Sample data:   ID     E1Date      E1Time   E2Date       E2Time   E3Date      E3Time    E4Date       E4Time    E5Date      E5Time  1      05JAN15    1715       05JAN15     2130      05JAN15    2355        06JAN15     0100       06JAN15    0300  2      15MAY11   1300                                                                          15MAY11    1523       15MAY11    1633 3      09FEB13                  09FEB13                  09FEB13    0800        09FEB13    1015       09FEB13    1145 4      18JUN17                                                                                      18JUN17                    18JUN17     data want; data have; length GoodDateTimeChron $8.; GoodDateTimeChron = 'NO'; DateTimeE1 = dhms(DateE1, 0, 0, TimeE1); /*create SAS datetime vars*/ DateTimeE2 = dhms(DateE2, 0, 0, TimeE2); DateTimeE3 = dhms(Date33, 0, 0, TimeE3); DateTimeE4 = dhms(DateE4, 0, 0, TimeD4); DateTimeE5 = dhms(DateE5, 0, 0, TimeE5); format DateTimeE1 DateTimeE2 DateTimeE3 DateTimeE4 DateTimeE5 datetime20.; /*this correctly flags row 1 by chronology, but none of the other rows due to missing data*/ if not missing(DateTimeE1) >=not missing(DateTimeE2) >=not missing(DateTimeE3) >=not missing(DateTimeE4) >=not missing(DateTimeE5) then GoodDateTimeChron = 'YES'; /*this doesn't directly flag for chronology, but it counts the number of distinct date values, including missing as a distinct value */ array event {5} DateE1 DateE2 DateE3 DateE4 DateE5 ; array new {5} $20 _temporary_; do _n_=1 to 5; new{_n_} = event{_n_}; end; call sortc(of new{*}); count = (new{1} > ' '); do _n_=2 to 5; if new{_n_} ne new{_n_-1} then count + 1; end; run; ... View more

Hello,   I need to divide a string variable into several component parts. All new variables will be string variables.   Sample data provided below:   OriginalVar     Component1    Component2    Component3 1702                 17                      02 1627M              16                      27                      M 1628SD            16                      28                      SD 18003               18                      003 1730S3             17                      30                      S3  1704/1              17                      04                      1   I used the following syntax:   data want;      data have;            Component1 = substr(originalvar,1,2); /*this variable is good*/            Rest=substr(originalvar,3); /*this variable contains everything from position 3 to end of OriginalVar*/            restAlpha=anyalpha(rest); /*identify position of 1st alpha in "REST"*/            restPunct=anypunct(rest); /*identify position of 1st punctuation in "REST"*/            restDigit=anydigit(rest); /*identify position of 1st digit in "REST"*/   if restAlpha = 0 and restPunct = 0 then Component2 = Rest;/*this syntax correctly creates Component2*/         else if restAlpha gt 0 and restPunct eq 0 then Component2 = substr(Rest,restDigit,restAlpha-restDigit);         else if restAlpha eq 0 and restPunct gt 0 then Component2 = substr(Rest,restDigit,restPunct-restDigit);     if restAlpha = 0 and restPunct = 0 thenComponent3 = ' '; /*Component3 is not properly generated*/         else if restAlpha gt 0 and restPunct eq 0 then Component3 = substr(Rest,restAlpha,3);         else if restAlpha eq 0 and restPunct gt 0 then Component3 = substr(Rest,restPunct,3);   end;   This syntax generates the following: OriginalVar     Component1    Component2    Component3 1702                 17                      02 1627M              16                      27                      M 1628SD            16                      28                      S         /*this is wrong - Component3 should = 'SD'*/ 18003               18                      003 1730S3             17                      30                      S        /*this is wrong - Component3 should = 'S3'*/ 1704/1              17                      04                      /         /*this is wrong - Component3 should = '1'*/     Any suggestions for how to fix my syntax?   Thanks so much!                            ... View more

I am working with date and time variables for 7 events that occur in a specific chronological order. As part of the QA process, I need to create a flag for a record in which any of the 7 events appear to be out of order. My data exists as a wide file (so information about all 7 events in the same row or observation).   Variables include:   reportID = unique report ID (1 report = 7 events) Event1_Date, Event2_Date, Event3_Date, Event4_Date, Event5_Date, Event6_Date, Event7_Date - SAS dates, format date9. Event1_Time, Event2_Time, Event3_Time, Event4_Time, Event5_Time, Event6_Time, Event7_Time -  SAS time, format time5.   I would prefer to NOT combine the date and time variables into a single datetime variable. While most reports have comprehensive date and time variables, there are some missing time variables (the date variables are 99.9% populated). Due to data errors, there are some incorrect date variables that appear to be both random and somewhat systematic (e.g., if the series of events spans 2 days that begins on the last day of a month and ends on the 1st day of the next month, there is a much greater likelihood of having an incorrect date variable for 1 of the events).  So while a date might be incorrect, the corresponding time appears to be correct.   I tried using the following syntax to begin creating a flag for events that might have incorrect/bad time variable values.   data want;      data have;     minute1=INTCK('minute',event1_time,event2_time);     minute2=INTCK('minute',event2_time,event3_time);     minute3=INTCK('minute',event3_time,event4_time);     minute4=INTCK('minute',event4_time,event5_time);     minute5=INTCK('minute',event5_time,event6_time);     minute6=INTCK('minute',event6_time,event7_time); run;   This syntax works - except for times that span midnight.   For example, if event1time = 23:50 and event2_time = 0:15 then minute1 = -1415 (when it should be 25).   Is there a way have the syntax work with times that span midnight, without combining the date and time variables into a single variable?   Any ideas will be most appreciated.       ... View more