Date:         Fri, 19 Oct 2001 13:32:59 -0400
Reply-To:     Dianne Rhodes <RHODESD1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Dianne Rhodes <RHODESD1@WESTAT.COM>
Subject:      Re: differences in Version 6.12 to Version 8 weighted univariate
Comments: To: Mike Rhoads <RHOADSM1@WESTAT.com>
Comments: cc: Ian Whitlock <WHITLOI1@WESTAT.com>,
          Brian Taaffe <TAAFFEB1@WESTAT.com>
Content-Type: text/plain; charset="iso-8859-1"

>
> However, my V6 Procedures Guide says that "The WEIGHT
> variable has no effect on the calculation of quantiles or
> extremes."  V8, however, does not contain this statement, and
> in fact documents how the weight variable is used in
> computing quantiles.
>
> So, check whether (1) in v6, your quantiles are the same with
> and without the WEIGHT statement, and (2) your quantiles
> without a WEIGHT statement in v8 match the v6 results.
>
Intuitively I knew you were right about this Mike.  The extreme values
matched but
not the quantiles.

When I run it in version 8 without the weight, I get the same numbers as in
Version 6 for the quantiles.
Version 6 produces the same quantiles with or without the weight.

> Please let me know whether or not this is correct.
>

Well, it is "correct" but I don't think it is what they wanted!

Bill's NOTE:
Also, PCTLDEF, in v8.2, is inoperative when using weights.
Does that mean that V 8.2 behaves like 6.12 when using weights?
Maybe I should just upgrade to 8.2?



        > Dianne Louise Rhodes
        > Sr. Systems Analyst
        > Westat
        >
        >