Calculate mean median IQR sum sumpct

[ybz12003]

Hello,

I have a dataset shown below. I would like to get mean/median/IQR/sum/sum_percent in each "Lab/Pham/Diag"  group, excluding the missing values.  Please let me know how to approach, thanks.

data Bill;  
	length caseid $10 Lab 3 Pham 4 Diag 4; 
	infile datalines delimiter=','; 
	input Caseid Lab Pham Diag;  
	datalines;                     
	EC1R00002,200,6000,5500,
	AC1R00012,100, , 7000,
	EC1R00013,50,3800, ,
	BH1U00185, ,4600,1500,
	EN1R01252,700,9000,2900,
	CY1K01251, ,500, , 
	BH1U99185,900,5600,8500,
	OP1R00252,360,9900,8900,
	PCHK01401,770,, , 
;   

[PaigeMiller · Accepted solution]

I don't know what you mean by "sum_percent"

Otherwise, this gets the job done:

proc means data=bill stackodsoutput mean median p25 p75 sum;
    ods output summary=summary;
    var lab pham diag;
run;

You will have to compute the IQR from the p75 and p25 values.

[ybz12003]

Lab (Sum_%)= Lab_sum / [Sum(Lab+Pha+Dia)]

[PaigeMiller]

---

@ybz12003 wrote:
Lab (Sum_%)= Lab_sum / [Sum(Lab+Pha+Dia)]

---

I'm not sure what this means. From the data you posted, the sums are

LAB 3080

PHAM 39400

Diag 34300

So are you asking for code to compute 3080/(3080+39400+34300) ??

[ybz12003]

Yes

[PaigeMiller]

proc means data=bill stackodsoutput mean median p25 p75 sum;
   ods output summary=summary;
   var lab pham diag;
run;
proc sql;
    create table want as select
    *,
    sum / sum(sum) as sum_pct
    from summary;
quit;

[ybz12003]

Is it a way to format the number to one digit after the decimal? 

[PaigeMiller]

Yes, assign the proper format to the variable.