And what about power transformations other than logarithms ( lamda =0.0 )
such as
Reciprocals -1.0
Square root .5
Reciprocal square root -.5
Etc. and for all values of lambda between -1.0 and 1.0
And of course GLS models where y1=y*w
Dave Reilly
Senior Vice President
Automatic Forecasting Systems
www.autobox.com
215-675-0652 (office)
215-353-7087 (cell)
-----Original Message-----
From: A UK-based worldwide e-mail broadcast system mailing list
[mailto:[log in to unmask]] On Behalf Of Lars Chi
Sent: Monday, April 27, 2009 6:26 PM
To: [log in to unmask]
Subject: Variance Stabilizing Transform - Queryþ
Hi,
I'd appreciate your opinion regarding the following. Suppose I apply a log
transform to my dependent variable and regress this transformed variable on
a set of independent variables. Now, the predicted values from this
regression are in log-scale. Is there any way I can obtain the predicted
values in the original scale. If I'm correct, simply exponentiation the
predicted values won’t do (I think as a consequence of Jensen's inequality,
these values will underestimate the true values).
Thanks in advance for your help!
L.C.
|