If you surveyed a whole population and made a chi-2 test and found a 
significant association between two categorical variables, you could 
interpret it as there must be a mechanism that creates an association 
between the two variables in the population. Since Y is not just 
randomly distributet independantly of  X.

Also if you did a Mann-Whitney test of differences in salaries between 
males and females, you could similarly interpret that there is a 
mechanism that prevents the ranks from being equally distributed 
between the two groups. It's possible to discuss the probability of 
different orderings.

For the t-test it's more difficult, I think since the t-distribution is a 
continuous one. But then I think you can use the line of reasoning that 
Robert Frick uses in an article in Behavior Research Methods, 
Instruments, & Computers a couple of years back. Frick uses the 
concept of PROCESS saying that there are inferences to 
PROCESSES and there are inferences to POPULATIONS. I think that 
the mechanisms I refer to above belongs to processes which have 
created the population at hand. And the chi-2 test and the Mann-
Whitney test could just as well refer to the process instead of the 
mechanism.

My suggestions is as I see them not in opposition to Radford Neal's but 
rather in line with them. Just put in a wiew of Robert Fricks discussion 
in his article. (Now I read a preprint, but I believe it is published.)

Rolf Dalin

> In article <[EMAIL PROTECTED]>,
> Tim Witort <trw7atixdotnetcomdotcom> wrote:
> 
> >I'm developing a report in an analysis program.
> >This report examines employee salaries - comparing
> >the salaries of men to those of women in a particular
> >job title in a particular company.  The goal is to
> >determine if the difference in their mean salaries
> >is statistically significant.
> >
> >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? 
> 
> Yes, because the "population" you are presumably interested in is not
> the population of actual current employees, but rather the population of
> possible employees and their salaries that would hypothetically result
> from continued appliation of the company's current employment and
> promotion policies into the indefinite future.  Only by making
> inferences about that hypothetical population can you conclude anything
> about the nature of these policies.
> 
>    Radford Neal
> 
> ------------------------------------------------------------------------
> ---- Radford M. Neal                                   
> [EMAIL PROTECTED] Dept. of Statistics and Dept. of Computer Science
> [EMAIL PROTECTED] University of Toronto                    
> http://www.cs.utoronto.ca/~radford
> ------------------------------------------------------------------------
> ---- . .
> =================================================================
> Instructions for joining and leaving this list, remarks about the
> problem of INAPPROPRIATE MESSAGES, and archives are available at: .     
>             http://jse.stat.ncsu.edu/                    .
> =================================================================

**************************************************
Rolf Dalin
Department of Information Tchnology and Media
Mid Sweden University
S-870 51 SUNDSVALL
Sweden
Phone: 060 148690, international: +46 60 148690
Fax: 060 148970, international: +46 60 148970
Mobile: 070 6740094, intnational: +46 70 6740094

mailto:[EMAIL PROTECTED]
http://www.ite.mh.se/~roldal/
**************************************************

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to