Donald Burrill said on 11/19/02 5:13 PM:

>It's not clear to me exactly what you mean by "a critical value
>approach", nor what other approach(es) you might have in mind to
>contrast it with.  But the consequent of that sentence is correct.
>(Although, if one is being pedantic, it is only true in the standardized
>metric, not in the metric of the variable observed;  in the latter
>metric "a significant distance from the parameter" would be the c.v.
>times the standard error of the observed statistic (the sample mean in
>the cases we've been discussing).)

Another approach, as compared to a critical value approach, is a 
probability approach. The text I'm currently using (Triola) distinguishes 
between 'traditional' and 'P-value' approaches to hypothesis testing and 
teaches both. 

The author has subsections devoted to each approach for the hypothesis 
testing sections. The traditional approach is exactly what you expect it 
to be, use the sample data to calculate a value for a test statistic and 
compare it to a critical value, basing your rejection decision on that. 
The P-value approach takes advantage of what computer technology gives 
you, a probability associated with the calculated test statistic. The 
probability is compared directly to alpha and the rejection decision is 
based on that. 

Of course, these approaches yield identical answers on problems. I like 
it because it bridges the gap between the old style and the current 
prevelance in use of computers. The traditional approach has advantages 
because it is graphically easy to illustrate. The P-value approach has 
value because it is more like what students may encounter in using a 
computer to calculate a stat and is increasingly what is reported in 
journals (less and less are we seeing p<.05 in lieu of p=.028, for 
instance). The linkage to probability is also more emphasized compared to 
the traditional approach where the probabilities are getting somewhat 
hidden behind veils of alpha, t, and z.

Now, if we can only get over the arbitrariness of the n<30 cut-off for 
use of t vs z and teach: use z when you know sigma and t when you don't. 
(Triola, as much as I like some of its choices, still retains this) <sigh>

Paul
.
.
=================================================================
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