Dennis Roberts seemed to utter in
news:[EMAIL PROTECTED]:
> At 02:31 PM 4/3/03 -0500, Rich Ulrich wrote:
>
>> >
>> > I have been directed to the t-test to gather this
>> > information. When I look at the t-test, however,
>> > it appears to be geared toward *estimating* the
>> > difference in the means of a population based on
>> > a *sample* of the population. Since I am using
>> > the entire population, can I still use the t-test
>> > to determine if the difference in the means is
>> > statistically significant? Is there another test
>> > that should be used instead?
>
>
> sort of agreeing with rich ... if there is a $1000 buck difference in
> average salaries ... that is what it is ... the notion of a
> "statistical" inference is irrelevant ... you could not attribute that
> to "sampling error" if the null were true
>
> since importance is not the same as statistical significance and, you
> don't really have an inference to make here ... doing something like a
> t test seems to me ... inappropriate
>
> if someone WANTS (insists) you do this ... because they have heard that
> to compare two groups you use a t test ... then they clearly do not
> understand what statistical inference is all about ...
That's exactly what is happening here. Not surprisingly,
it is the federal government who insists on this
analysis. :^)
> now, if you are assuming that your males and females are but samples
> of all the years and salaries that you have had or could have (ie, target
> population is much much larger) ... perhaps THEN ... some inferential test
> might be ok
I found Radford Neal's reply to this the most logical
support for doing the t-test in this scenario. When
looked at from his perspective (similar to what you said
here) it can be supported. I guess I have to look at it
in this way to feel I'm not misapplying this analysis.
-- TRW
_______________________________________
My e-mail: t r w 7
@ i x . n e t c o m . c o m
_______________________________________
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
