[EMAIL PROTECTED] (Jon Cryer) wrote in message
news:<[EMAIL PROTECTED]>...
> The TI-83 uses the Saitherwaite (spelling?) approximation for the degrees
> of freedom.
> This is not the same as using the conservative one less than the smaller
> sample size.
>
> Jon Cryer
Thanks, I guess that is the answer I was looking for; I had never
heard of the Sattherwaite approximation before, but I found the
formula on the Web. (In my example the TI-83 says df = 8.09.) I cannot
find it mentioned in the TI-83 owners manual nor in any of the
introductory texts I have, so I suppose there must be some reason why
the authors of these texts choose the more conservative approach (even
though they usually have sections explaining to the students how to do
the tests on the TI-83). In any case, I suppose that I will have to
explain to the the class that the TI computes the DF differently from
the textbook in the case of the non-pooled t-test or t-interval, and
that could possibly change the outcome of a t-test.
[EMAIL PROTECTED] wrote:
=If you do not have access to a package like SPSS which is outstanding
in
=its human factors, (clarity, tutorials, stat coach etc.), you might
=google web calculators.
I don't understand what you mean by "google web calculators". I
wish I had the computer lab time available and the time in the course
to teach them SPSS or Minitab.
[EMAIL PROTECTED] wrote:
=- of course, you are not supposed to make the choice of
=test on this ad-hoc basis.
=
=To be more explicit: according to a couple of
=articles that I read (and I agreed on this point), it is a BAD
=practice to 'condition' your choice. You are not
=supposed to let the test of variances determine
=which t-test to believe.
=
I didn't mean to go into that, but actually the text I am using
does that (chooses the pooled or non-pooled t-test based on the test
of variances), although most of the other texts I have seen do not.
In my class we don't even cover the test for equality of variances.
For an introductory course for students who are math-anxious, I try
not to delve into too much of the theoretical issues, so I tell the
class to use the non-pooled t-test unless instructed otherwise.
Howard Wachtel
[EMAIL PROTECTED]
>
> At 12:13 PM 1/6/2003 -0800, you wrote:
> >I am teaching introductory statistics even though statistics is not
> >really my specialty. I encourage my students to use a TI-83
> >calculator(even though most of them cannot afford one). However there
> >is one statistical procedure that always comes up with a different
> >result when I use the TESTS menu on the TI-83 versus using a
> >t-distribution table and the formula in the textbook, namely a
> >non-pooled t-confidence interval for the difference between two means.
> > (Every other statistical procedure, e.g. a pooled or non-pooled
> >t-test, or a pooled confidence interval, comes out the same when I do
> >it either way.)
> >
> >For example: x1bar = 349, x2bar = 383 (sample means), s1 = 19.6, s2=
> >39.5 (sample standard deviations), n1 = 10, n2 = 7 (sample sizes),
> >find a 95% confidence interval for mu1 - mu2 (difference in population
> >means). When I use the formula in the textbook with critical t =
> >2.447 (df = 6, using one less than the smaller sample size) from the
> >table, I get 349 - 383 +/- 2.447*sqrt[19.6^2/10 + 39.5^2/7] which
> >comes out to the interval (-73.56, 5.56). However, when I put the
> >same information into the TI-83 (option "0" on the TESTS menu) and
> >choose "no" for pooled, it gives me the interval (-71.21, 3.21). If I
> >do a t-test with the same data, it comes out the same either way.
> >What am I doing wrong?
> >
> >Thank you.
> >
> >Howard Wachtel
> >Delaware Valley College
> >[EMAIL PROTECTED]
> >.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
