When SAS applies QMETHOD=P2 in PROC MEANS

[Phil_NZ]

Hi all, 

When running the proc means to a dataset, I face a normal problem that

WARNING: A shortage of memory has caused the quantile computations to terminate prematurely for QMETHOD=OS. Consider using 
         QMETHOD=P2.

In particular, while all of the summary characteristics work, only median spot the "." in my result with the normal PROC MEANS

proc means data=merge_treat_con n nmiss mean median std min max skew kurt;
run;

When I use the option QMETHOD=P2

proc means data=merge_treat_con n nmiss mean median std min max skew kurt qmethod=p2;
run;

And a quick comparison between these two Tables of the result, it seems that they only differ in the "median" column, so I deem that option qmethod=p2 only affects the unsolvable column while using proc means without this option, other results would remain the same.

Could you please let me know if I fell into any fallacy?

Warm regards.

[Reeza · Accepted solution]

QMETHOD affects only the percentiles. So anything related to quartiles or percentiles are affected. In the statistics you've chosen to display that would only be the median but the first and third quartiles would also be affected.

[Phil_NZ]

Hi @Reeza 

Thank you for your reply, so it also means that the 1st and 99th percentile also been affected, am I correct? Just want to cross-check if it is the case.

Regards.

[Reeza]

Yes, they can be affected, they may not be though that's the catch.

[Phil_NZ]

Hi @Reeza 

So, you mean, they either can or can't be affected, is it your point?

Warm regards,

[ballardw]

---

@Phil_NZ wrote:

Hi @Reeza 

Thank you for your reply, so it also means that the 1st and 99th percentile also been affected, am I correct? Just want to cross-check if it is the case.

Regards.

---

The documentation says:

When QMETHOD=P2, reliable estimations of some quantiles (P1,P5,P95,P99) might not be possible for some data sets.