Date:         Fri, 7 Sep 2007 11:11:29 -0500
Reply-To:     Warren Schlechte <Warren.Schlechte@TPWD.STATE.TX.US>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Warren Schlechte <Warren.Schlechte@TPWD.STATE.TX.US>
Subject:      Re: PROC NPAR1WAY One-Sided Tests
Content-Type: text/plain; charset="us-ascii"

In some cases, isn't it the ranks that are important?

In those cases, use PROC Rank to rank the data, then get the sum of the
ranks to get an idea of which treatment has the higher/lower ranked
values.

Warren Schlechte

-----Original Message-----
From: Jim Groeneveld [mailto:jim4stat@YAHOO.CO.UK] 
Sent: Friday, September 07, 2007 9:43 AM
Subject: Re: PROC NPAR1WAY One-Sided Tests

Hi Paige,

Well, that is not so difficult. Just look at the median, not the mean
(you
do it nonparametric don't you? Though in the output of NPAR1WAY I see
means)
values of your effects; are these in the expected direction? If these
are
not included in the standard output it must be possible to caugh them up
in
some other way. And you know the direction that you expect. But maybe
the
PROC could be more clear about this. Yet it is possible to interprete
the
output correctly.

Regards - Jim.
--
Jim Groeneveld, Netherlands
Statistician, SAS consultant
home.hccnet.nl/jim.groeneveld


On Fri, 7 Sep 2007 05:08:37 -0700, Paige Miller <paige.miller@KODAK.COM>
wrote:

>On Sep 7, 4:38 am, jim4s...@YAHOO.CO.UK (Jim Groeneveld) wrote:
>> Hi Paige,
>>
>> It may seem somewhat unclear, but if you look carefully at the
example
>> output you'll always see a one-sided p-value being half of the
two-sided
>> p-value, which is indeed correct. E.g. if your alpha limit is .05 you
have
>> .025 at each end (of the normal curve) with two-sided testing, but
.05 at
>> just one end with one-sided testing. That means that at the right
one-sided
>> end some result might be significant at the one-sided level, but not
at the
>> two-sided level, e.g. if the p-value is .04. Well, even this may not
be
>> explained quite clearly (it can be explained in several ways), but
what I
>> want to say is that some effect in the expected direction is "more
>> significant", reaches significance earlier (with a smaller effect in
the
>> expected direction) with one-sided testing than with two-sided
testing. This
>> has been presented by PROC NPAR1WAY by halving the two-sided p-value.
>
>This part makes sense.
>
>> You have to interprete this in the one-sided case as:
>> If the effect found with the sample is in the expected direction then
take
>> the one-sided p-value and draw your conclusions.
>> If the effect found with the sample is in the opposite direction then
your
>> effect is not significant at all and you can ignore the p-values at
all.
>> The procedure does not know about your hypothesis, so it just offers
a
>> one-sided p-value for the case the found effect is in the expected
direction.
>
>I still cannot fathom this part. How does one determine if "the effect
>found with the sample is in the expected direction"? Let us suppose
>that my alternate hypothesis is mean group 1 < mean group 2. How can I
>tell from the output which direction the sample points? The SAS output
>doesn't tell me (at least I can't find it in the docs) which of my two
>groups is considered group 1, and which of my two groups is group 2.
>Furthermore, SAS doesn't tell me what the z value represents ... does
>it represent group 1 minus group 2, or group 2 minus group 1. Can you
>see why this is unclear?
>
>--
>Paige Miller
>paige\dot\miller \at\ kodak\dot\com