On 13 Aug 2003 09:10:25 -0700, [EMAIL PROTECTED] (Jay Warner) wrote:

>I like Don Burrill's suggestion.  It can be done easily.  Furthermore, if
>you are not reasonably familiar with regression methods, why would you
>seek out a second or 3rd level text?  If you aren't reasonably sure that
>your simplifying data-related assumptions will lead you astray in terms
>of precision of predictions, then a 'complete' treatment is probably a
>waste of intellectual energy.  Which is in itself a most precious
>resource.
>
>to quote a number of statisticians, the first thing you do is PLOT YOUR
>DATA!  Then look at it.  IF and only if it appears to follow an
>exponential decay form, then a transformation will make the plot look
>straight, and you can consider fitting a straight line to it and making a
>prediction from there.  Based on the statistical precision you can
>demonstrate at that point.

My hypothesis is based on widespread claims in the literature that
memory does follow an exponential decay curve. I want to test that
concept in specific situations.

>If I were to undertake this type of study, the first thing I would do is
>even before that.  I would ask how you intend to measure 'memory' and
>'memory loss.'  I would ask for operational definitions of the terms
>thrown around, and then would ask carefully whether the thing which you
>measure (such as % of items recalled) actually indicates what you think
>it does (memory loss?).  When you could make clear, supportable
>statements about what you would measure, and how, then I would say to
>collect some data and see what the plot looked like.

Yes, of course. All good and proper procedural steps. 

>And good luck on your project.  I'd add one other thing, but I forgot
>what it was :)

See, you need my memory study!

-- 
Spam sink email address, sorry
.
.
=================================================================
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