One-tailed t-test

[Sean M. Connery]

I am running SPSS ver 7.5 for Windows. Does anyone know how to do a
ONE-tailed, independent sample t-test.

Thanks,

Sean M. Connery

(not the rich one)

[Jan Radder]

For one-tailed testing you can take half the two-tailed
p-value. You then can conclude there is a statistical
significant difference between the two samples if this value
is equal to or less then the risk you want to take for
drawing a false conclusion AND the difference between the
means is in the hypothesized direction.

Jan Radder

Jan Radder -- jj.r...@pg.tno.nl

[Edward Torpy]

Hi Sean,

When you run a independed sample t-test in SPSS 7.5, clicking on the
"Options..." button allows you to specify any confidence interval you
wish. Thus, a "ONE-tailed" t-test with an alpha of .05 would have a
confidence interval of 97.5%.

- Ed
-----------------------------------------------------------------------
Edward Torpy Technical Marketing Specialist SPSS Inc.
Phone: (312) 494-3289 Internet: eto...@spss.com Fax: (312) 329-3690
-----------------------------------------------------------------------

[Richard F Ulrich]

Edward Torpy (eto...@spss.com) wrote:

: When you run a independed sample t-test in SPSS 7.5, clicking on the

: "Options..." button allows you to specify any confidence interval you
: wish. Thus, a "ONE-tailed" t-test with an alpha of .05 would have a
: confidence interval of 97.5%.

- Whence 97.5% ?

In my imagination, a one-tailed test at 5%
looks just like a 10% test that is two-tailed,
except that you have to chop off the unwanted tail
(which represents the impossible outcome).

--
Rich Ulrich, biostatistician wpi...@pitt.edu
http://www.pitt.edu/~wpilib/index.html Univ. of Pittsburgh

[John A. Grossbohlin]

Edward Torpy <eto...@spss.com> wrote:

>When you run a independed sample t-test in SPSS 7.5, clicking on the
>"Options..." button allows you to specify any confidence interval you
>wish. Thus, a "ONE-tailed" t-test with an alpha of .05 would have a
>confidence interval of 97.5%.
>

Is there any reason this would not apply to a matched pair t-test?

John

[David Nichols]

In article <6d23r8$f...@usenet.srv.cis.pitt.edu>,
Richard F Ulrich <wpi...@pitt.edu> wrote:

>Edward Torpy (eto...@spss.com) wrote:
>
>: When you run a independed sample t-test in SPSS 7.5, clicking on the
>: "Options..." button allows you to specify any confidence interval you
>: wish. Thus, a "ONE-tailed" t-test with an alpha of .05 would have a
>: confidence interval of 97.5%.
>

> - Whence 97.5% ?
>
>In my imagination, a one-tailed test at 5%
>looks just like a 10% test that is two-tailed,
>except that you have to chop off the unwanted tail
>(which represents the impossible outcome).
>
>
>--
>Rich Ulrich, biostatistician wpi...@pitt.edu
>http://www.pitt.edu/~wpilib/index.html Univ. of Pittsburgh

I think what Ed meant here was that you could look at an alpha/2 CI to
see if it contained 0 in order to get the same conclusion as a one-sided
test at alpha, which is correct. If you want a one-sided p-value, you
of course simply cut the two-sided one in half if the mean difference is
in the predicted direction.

--
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.

[David Nichols]

In article <34f7892a...@news.ulster.net>,
John A. Grossbohlin <gros...@ulster.net> wrote:

>Edward Torpy <eto...@spss.com> wrote:
>
>>When you run a independed sample t-test in SPSS 7.5, clicking on the
>>"Options..." button allows you to specify any confidence interval you
>>wish. Thus, a "ONE-tailed" t-test with an alpha of .05 would have a
>>confidence interval of 97.5%.
>>
>

>Is there any reason this would not apply to a matched pair t-test?
>
>John

No, the same principle would apply.