I suggest looking at the books Tukey co-authored, for help 
with basic ideas of scoring and scaling.  Others might have
additional suggestions.


On 18 Jun 2003 15:14:24 -0700, [EMAIL PROTECTED] (Dianne Worth)
wrote:

> It is my understanding that if the data distribution is positively 
> skewed , one can take a log of the data.  

One CAN, and it might help.  And it *might*  be a totally
ignorant, wrong-headed thing to do.  It is not surely not
automatic.  One might also take the square root, or the
reciprocal, or form groups out of scores.  Or, one might
suffer no harm from leaving them alone; or, one might 
suffer harm but *still*  be best-advised  to leave them alone.

>                                    If it's negatively skewed, 
> one should 'flip' it by adding 1 to the largest value, and a series of 
> steps that are probably best explained as follows:

It is certainly a good idea to *consider*  the meaning of 
reversing the scores, especially when you have scores (as 
you seem to) that are integer scale-scores.  On the other 
hand, as *automatic*  advice, that is totally stupid, starting
with "adding 1".    And if they are scale-scores, the square
root is more likely to suit, than the log.

For negative integers, the obvious change is to add 
enough so the bottom is zero, or to reverse them.  - Where
do the numbers come from? - that probably suggests 
which makes more sense for making them positive.

Where do the numbers come from?  Are they so abstract
that there is never any harm done, to their meaning, by 
taking any arbitrary transformation?  If you don't know anything
more than "order", it might be wise to rank-transform them.
(I usually consider ranking a pretty weak step to take -- It
is an admission that the data-collection was poor enough,
we  can not extract any information beyond  the  'ordering'.)

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization."  Justice Holmes.
.
.
=================================================================
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