Different results between proc surveyselect and %BOOT

CHELS — Wed, 25 Jul 2018 19:24:28 GMT

I was trying out both surveyselect and %boot on the same set of data, thinking they will give me the same results;however, I did not get that. The 95% PTCI from two procedures are different and here are the code:

For %boot:

%macro analyze(data=,out=);
proc means noprint data=&data;
%bystmt;
var x;
output out=&out(drop=_type_ _freq_) skew=_Skewness;
run;
%mend;

%boot(data=RandInt,
samples=5000,
random=12345,
chart=0,
stat=_Skewness,
alpha=0.05,
print=1);

proc print data=bootstat noobs;
id method n;
var value bootmean bias stderr biasco alcl aucl;
run;

%bootci(PCTL);

For survey select:

%let NumSamples = 5000;
proc surveyselect data=RandInt NOPRINT seed=12345
out=BootSSFreq(rename=(Replicate=SampleID))
method=urs
samprate=1
OUTHITS
reps=&NumSamples;
run;

proc means data=BootSSFreq noprint;
by SampleID;
freq NumberHits;
var x;
output out=OutStats skew=Skewness;
run;

proc univariate data=BootSSFreq noprint;
var Skewness;
output out=Pctl
pctlpre =CI95_
pctlpts =2.5 97.5
pctlname=Lower Upper;
run;

proc print data=Pctl noobs; run;

Re: Different results between proc surveyselect and %BOOT

Reeza — Wed, 25 Jul 2018 19:47:12 GMT

1. How far off are they? Since this is a simulation they should be really close but not necessarily exact.

2. Look at the VARDEF options in PROC UNIVARIATE and check the note about the df and n-1 denominator in the %boot macro, if that's applicable here.

Re: Different results between proc surveyselect and %BOOT

CHELS — Wed, 25 Jul 2018 21:05:35 GMT

They are the same up to 1 decimal place. I guess I have to choose one of the two to use then. THANK YOU!

topic Re: Different results between proc surveyselect and %BOOT in Statistical Procedures

Different results between proc surveyselect and %BOOT

Re: Different results between proc surveyselect and %BOOT

Re: Different results between proc surveyselect and %BOOT