How do you keep variables in the new dataset you create using proc means (not manipulating them)?

[Flexluthorella]

data Hivdata;
merge learn.demogdemo (keep=patient_id male death arvstart) learn.visitdemo (keep=patient_id weight apptdate);
by patient_id;
run;

*identify min weight prior to ARTstart by seperating records where appt date is less than or equal to arvstart to find min weight by id;

proc means data=work.Hivdata min noprint;
class patient_id;
var weight male death;
where apptdate le arvstart;
output out=minweight
min=minWt;
run; 

This code is outputing the new var minWt but the other variables like male and death are not being keep. I also need to do this with max weight and compare difference.

[unison]

If you want your other variables you need to add them into the class statement - be careful though as this can change the result of the minimum just based on patientid. What you can do is merge the result of proc means back to your original set.

-unison

[Flexluthorella]

How will it changes the results? I need the absolute min per patient ID

[Flexluthorella]

when I ran the var in the class statement, there were missing data spots. What do you mean by adding proc means results to original data set? How would I do that

[PaigeMiller]

---

@Flexluthorella wrote:

data Hivdata;
merge learn.demogdemo (keep=patient_id male death arvstart) learn.visitdemo (keep=patient_id weight apptdate);
by patient_id;
run;

*identify min weight prior to ARTstart by seperating records where appt date is less than or equal to arvstart to find min weight by id;

proc means data=work.Hivdata min noprint;
class patient_id;
var weight male death;
where apptdate le arvstart;
output out=minweight
min=minWt;
run; 

This code is outputing the new var minWt but the other variables like male and death are not being keep. I also need to do this with max weight and compare difference.

---

If I am understanding the question properly (which I may not be, can you clarify or explain in more detail what you want) you need ask for statistics for variables MALE and DEATH in the OUTPUT statement as well.

Example:

output out=minweight min(weight)=minweight mean(male death)=;

[Flexluthorella]

I need the min weight per patient_id. the var male and death are numeric. male is gender and values include 1 for male 0 for female. Death values include 1 for deceased 0 for not deceased. I do not need the min or any stats for these.

[Tom]

---

@Flexluthorella wrote:
I need the min weight per patient_id. the var male and death are numeric. male is gender and values include 1 for male 0 for female. Death values include 1 for deceased 0 for not deceased. I do not need the min or any stats for these.

---

So it sounds like those will be constant values across all observations for a particular person.

You can either include them in the CLASS statement instead of the VAR statement. (make sure to add NWAY option to the PROC statement).

Or if you include them in the VAR statement then calculate a stat on them.  Like MIN or MAX.

[novinosrin]

Hi @Flexluthorella  Good morning, A sample of your input dataset and the expected output for the input will help determine what's the best approach i would think.  Your description gives me leaning towards proc sql unless your variable male and death are actually analysis variables.

[Flexluthorella]

Male and death need to be in the final table but not analyzed. I am creating two datasets with minwt and maxwt then merging these two together. My final output should have the delta weight where delta is gt 10 lbs and the tables will have only the proportion of deceased (death=1 only) or males vs females. final table is most likely proc freq with cum percentage.

[novinosrin]

ok @Flexluthorella  Thanks. May I/we(on behalf of othersw too) request you to post the values of what you have?

Is it something like the below?

Patientid  Gender Death Weight

A01            M           0       40

A02            M           1        60

A03            F            0        65

A04           F              1        55

[Flexluthorella]

See attached for initial data set. there are several wt values for each patient_id. I need the min and the max for each pat ID. The original file was too large to upload.

[novinosrin]

How about?

proc sql;
select patientid, male, deceased ,min(weight) as min_wt
from your_input
group by patientid
having weight= min_wt;
quit;

[Flexluthorella]

Can I do the same for maxweight? Can I do this in one proc sql?

[novinosrin]

yes you can if you want something like the following this to get two records i.e 1 for min 2 for max in the having filter

proc sql;
select patientid, male, deceased ,min(weight) as min_wt,max(weight) as max_wt
from your_input
group by patientid
having weight= min_wt or weight=max_wt;
quit;

[Flexluthorella]

There is an error with this one as well. The query does not give me the output I need