Dennis Roberts wrote:
>
> in any analysis ... whether with 2 or 50 or 400 ... you have to ask
> yourself: what am i doing this for?
>
> are you just trying to summarize and describe the data in front of you?
> are you using the data as some estimate of a larger population of data?
>
> if it is the first purpose and, you have only several values ... you just
> lay them out for sore eyes to see ... and, even sore eyes will see what the
> data are like ... if you have 50 or 400 ... condensing methods of
> summarizing the data might be useful ... to sore OR good eyes
>
> if it is the second purpose ... then the issue becomes: do i have enough
> data (that were gathered in a good way) so that my estimate of what the
> population looks like ... will be decently accurate? then n becomes important
With small n, one must often also ask: are there enough data to make
you justified in believing in any model for which the standard deviation
is a good measure of spread?
For instance, it is often *not* a natural choice for strongly bimodal
distributions, skewed distributions, or heavy-tailed distributions.
-Robert Dawson
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
