Re: Negative R2
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Sun, 04 Feb 2007 23:13:47 -0500
On 3 Feb 2007 15:27:18 -0800, "Deep" <Deepangi.Pandit@xxxxxxxxx>
wrote:
Hello Group:
I am learning statistics and new to its subtleties. I am using partial
least square (PLS) method to develop my model. It is suggested in
literature to check the stability of model by plotting Predicted vs.
Actual and Actual vs. Predicted and for both also check fit line
through the origin.
That sounds like a suggestion for a graphical technique.
It only makes some sense for them to intercept near zero
if there is *reason* for it, such as, predicting quantities,
where the null amount would be zero, and the prediction is
linear.
However I get negative R2 (rsquare) for line
fitted through origin for plot y= actual and x= predicted.
- But *this* sounds like you provided a further regression
of something-or-other, which I have never heard of.
Could you
please explain the reason? I understand that this line does not fit in
a linear fashion and may be it is curved but I would appreciate
insight from the group members.
Check again what you were reading about how to check the
stability of the model. I think you are misusing it.
--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.
- References:
- Negative R2
- From: Deep
- Negative R2
- Prev by Date: Re: Testing Hypotheses
- Next by Date: Re: Negative R2
- Previous by thread: Re: Negative R2
- Next by thread: Re: Negative R2
- Index(es):
Relevant Pages
|
