At the risk of sounding totally ignorant, what is a
"lamba transform," or where could I read about it?
DW
--- Jay Warner <[EMAIL PROTECTED]> wrote:
> No, and maybe yes.
> 
> What you are doing is transforming the x axis so
> that the resultant
> distribution appears as a Normal.  If it works :)
> 
> A 'positively skewed' distribution is not
> infrequently something like
> a log-Normal.  Thus, a straight log of the x values
> gives a nice
> Normal upon transformation.
> 
> Look at the picture, already!  (sorry.  I get
> frustrated...)
> 
> the negatively skewed distribution can be
> transformed also.  Either
> 'flipping it, as you suggest, or perhaps by using
> Taguchi's 'omega'
> transform.  (It has another name in 'western'
> literature, but I don't'
> know what - arc-sine?)
> 
> x' = ln(x/(w-x))
> 
> where x is the original x value, w is a value
> slightly in excess of
> the highest possible x.  It could be ln or log,
> either way, and x' is
> the new x scale value.
> 
> this works nicely, very nicely.  You can get a shift
> in the resultant
> distribution, by adjusting w if you like. 
> Interpretation leaves
> something to be desired.....
> 
> If the x scale values are less than 0, you need to
> be careful to
> ensure that the ln is performed on positive values,
> is all.
> 
> If you can justify the transformation based on the
> technology you are
> playing with, so much the better.  If you justify it
> on the basis that
> the resultant distribution is 'more' Normal, you
> might as well go for
> a lambda transform, and be done with it.
> 
> Cheers,
> Jay
> 
> Dianne Worth wrote:
> 
> > It is my understanding that if the data
> distribution is positively
> > skewed , one can take a log of the data.  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:Original values
> > are 1-8.Add 1 to 8 =
> 9.FLIPX=9-XLOGFLIPX=LOG(FLIPX)LOGX=.903089 -
> > LOGFLIPXwhere .903089 is the log of the original
> largest value
> > (8). My question: What if my positively skewed
> data all  have
> > negative values, say from -1 to -7?  Does the same
> logic/procedure
> > still hold?Thanks,DW
> >
>
---------------------------------------------------------------------
> > Do you Yahoo!?
> > SBC Yahoo! DSL - Now only $29.95 per month!
> 
> --
> Jay Warner
> Principal Scientist
> Warner Consulting, Inc.
> 4444 North Green Bay Road
> Racine, WI 53404-1216
> USA
> 
> Ph: (262) 634-9100
> FAX: (262) 681-1133
> email: [EMAIL PROTECTED]
> web: http://www.a2q.com
> 
> The A2Q Method (tm) -- What do you want to improve
> today?
> 
> 
> 
> 
> .
> .
>
=================================================================
> Instructions for joining and leaving this list,
> remarks about the
> problem of INAPPROPRIATE MESSAGES, and archives are
> available at:
> .                  http://jse.stat.ncsu.edu/        
>            .
>
=================================================================


__________________________________
Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!
http://sbc.yahoo.com
.
.
=================================================================
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