topic Re: SQL observation summary question in SAS Programming

SQL observation summary question

rajd1 — Fri, 29 May 2020 04:02:23 GMT

Hello there,

I have been trying to work on this problem with data manipulation using SQL. Here is my program and my data set:

ID	Cycle	Delay	Reduction	DrugName	Regimen	Dose
1	1	None	None	Carb	x,y	750
1	1	None	yes	Pem	x,y	1015
1	2	None	None	Pem	x,y	1015
1	2	None	None	Carb	x,y	705
1	1	yes	None	Durv	x	940
1	2	None	None	Durv	x	940
1	3	None	None	Durv	x	930
1	4	yes	None	Durv	x	910
1	5	None	None	Durv	x	910
1	6	None	None	Durv	x	910
1	7	None	None	Durv	x	940
1	8	None	None	Durv	x	945
1	9	None	yes	Durv	x	920
1	10	None	None	Durv	x	915
1	11	None	None	Durv	x	915
1	12	Unknown	None	Durv	x	920
1	13	None	None	Durv	x	920
1	14	None	None	Durv	x	920
1	15	None	yes	Durv	x	925
1	16	None	None	Durv	x	910
1	17	None	None	Durv	x	890
1	18	None	None	Durv	x	875
1	19	None	None	Durv	x	870
1	20	None	None	Durv	x	850
1	21	None	None	Durv	x	850
1	22	None	None	Durv	x	845
1	23	None	None	Durv	x	860
1	24	None	None	Durv	x	855
2	1	None	None	Pacl	x,z	95
2	1	None	None	Carb	x,z	245
2	2	None	None	Pacl	x,z	95
2	2	None	None	Carb	x,z	225
2	3	None	None	Pacl	x,z	95
2	3	None	Other	Carb	x,z	225

Here is my program:

Proc sql;
Create table Drug_Clean AS
select distinct (DrugName),
ID as ID, max(cycle) as maxcycle, min(cycle) as mincycle, Count(Dose) as numberdoses, First(dose) as Firstdose, mean(Dose) as MeanDose, last(Dose) as LastDose
from L_line1
Group by SubjectID;
quit;

Here is the final dataset that i would like to have the following:

It should have all the drugnames listed by id: For exampl ID #1 has carb, pem and durv. I want to display as follows:

ID	Drug	MaxCycle	MinCycle	countdose	Avgdose	FirstDose	LastDose	Reduction	Delay
1	Durv	24	1	24	902.7	940	855	yes	yes

Any help is greatly appreciated!!!

Re: SQL observation summary question

ChrisNZ — Fri, 29 May 2020 03:54:20 GMT

Why is there no Carb in the output?

Re: SQL observation summary question

rajd1 — Fri, 29 May 2020 03:59:59 GMT

Sorry, Yes I want it to display all the carb and pem as well. Basically all the drugnames by ID.

Thanks

Re: SQL observation summary question

ChrisNZ — Fri, 29 May 2020 04:34:12 GMT

Like this?

proc sql;
  create table WANT as
  select unique ID
         , DRUGNAME
         , MAXCYCLE
         , MINCYCLE
         , NUMBERDOSES
         , MEANDOSE   
         , max(FD) as FIRSTDOSE
         , max(LD) as LASTDOSE
  from (
    select unique ID
         , DRUGNAME
         , max (CYCLE) as MAXCYCLE
         , min (CYCLE) as MINCYCLE
         , count(DOSE) as NUMBERDOSES
         , mean(DOSE)  as MEANDOSE
         , DOSE*(CYCLE=calculated MINCYCLE) as FD
         , DOSE*(CYCLE=calculated MAXCYCLE) as LD
    from HAVE
    group by ID, DRUGNAME
  )
  group by ID, DRUGNAME
  order by ID, DRUGNAME
  ;
quit;

Re: SQL observation summary question

ed_sas_member — Fri, 29 May 2020 08:02:27 GMT

Hi @rajd1

Here is a way to achieve this.

Could you please clarify the rule to compute delay and reduction?

In the following program, I have set reduction and delay to "yes" if there was at least one record equal to "yes", and "none" in other cases.

However, I can see that there is a value "Other" for patient 2 at cycle 3. How would you like to handle this case?

Best,

proc sql;
	select distinct ID, DrugName, MaxCycle, MinCycle, countdose, Avgdose, max(FirstDose) as FirstDose, max(LastDose) as LastDose, Delay, Reduction
	from (select ID, DrugName,
		   max(cycle) as MaxCycle,
		   min(cycle) as MinCycle,
		   count(dose) as countdose,
		   mean(dose) as Avgdose format=8.1,
		   dose*(cycle=calculated MinCycle) as FirstDose,
		   dose*(cycle=calculated MaxCycle) as LastDose,
		   case when sum(lowcase(delay)="yes") > 1 then "yes" else "none" end as Delay,
		   case when sum(lowcase(reduction)="yes") > 1 then "yes" else "none" end as Reduction /*what if "Other" ?*/
	 	 from L_line1
		 group by ID, DrugName)
	group by ID, DrugName	
	order by ID, DrugName;
quit;

Re: SQL observation summary question

rajd1 — Fri, 29 May 2020 21:10:57 GMT

This worked GREAT! The delay variable has 3 categories: none, yes, and unknown. I would like to make these as categories for the delay variable as well.

basically i want to be able to add the 3rd category to the delay variable.

Further i have some dates and was trying to find the difference between dates.

i was working with and here is the code:

proc sql;
create table clean_Drug as
select unique SubjectID
, DRUGNAME
, REGIMEN
, STARTCYCLE
, ENDCYCLE
, CYCLEDAYS
, STARTLOT
, ENDLOT
, MAXCYCLE
, MINCYCLE
, AVGCYCLE
, MEDIANCYCLE
, NUMBERDOSES
, MEANDOSE
, max(FIRSTDOSE) as FIRSTDOSE
, max(LASTDOSE) as LASTDOSE
from (
select unique SubjectID
, DRUGNAME
, REGIMEN
, min (LotStart) as STARTLOT format MMDDYY10.
, max (LotEnd) as ENDLOT format MMDDYY10.
, min (AdminDate) as STARTCYCLE format MMDDYY10.
, max (AdminDate) as ENDCYCLE format MMDDYY10.
, max (CYCLE) as MAXCYCLE
, sum (ENDCYCLE,-STARTCYCLE) as CYCLEDAYS
, min (CYCLE) as MINCYCLE
, mean (CYCLE) as AVGCYCLE
, median(CYCLE) as MEDIANCYCLE
, count(DOSE) as NUMBERDOSES
, mean(DOSE) as MEANDOSE
, DOSE*(CYCLE=calculated MINCYCLE) as FIRSTDOSE
, DOSE*(CYCLE=calculated MAXCYCLE) as LASTDOSE
from L_Line1
group by SubjectID, DRUGNAME)
group by SubjectID, DRUGNAME
order by SubjectID, DRUGNAME
;
quit;

I got the following error:

ERROR: Function SUM requires a numeric expression as argument 1.
ERROR: The following columns were not found in the contributing tables: ENDCYCLE, STARTCYCLE.

Thanks so much!

Re: SQL observation summary question

rajd1 — Fri, 29 May 2020 20:48:27 GMT

This worked GREAT! Thanks so much!
Quick question: say i wanted to have one row for each ID and have drug and cycle listed as variables like Drug1, Drug2, and so on. How can i modify this code? I tried a few things and it didn't work. I am soon realizing the power of SQL. This is a great community and i have been using this everyday to learn:)

Re: SQL observation summary question

ChrisNZ — Fri, 29 May 2020 23:58:51 GMT

SQL is not very suited to the first task (first and last row), and even less suited to what you want now (create an unknown number of columns). 1. These 2 tasks are best performed in a data step 2. The structure you are requesting is ill-suited to any kind of processing. It's much better to keep the data vertical.