| |
 |
|
|
Science Forum Index » Statistics - Math Forum » principal component regression...
Page 1 of 1
|
| Author |
Message |
| Seth Myers... |
 Posted: Mon May 05, 2008 11:12 am |
|
|
|
Guest
|
Simple question...
If I have a set of variables and perform a PCA to create from them a few non-correlated variables, will the slope coefficients for these principal components (in a linear model) be the same no matter how (order,etc) they are entered into the model?
For instance, I would like it if after the following two functions were fit, B1 from the first and second function would be the same.
y = int + B1*PC1 + B2*PC2 + B3*PC3
y = int + B1*PC1
thanks,
seth |
|
|
| Back to top |
|
| Ray Koopman... |
| Posted: Mon May 05, 2008 1:24 pm |
|
|
|
Guest
|
On May 5, 2:12 pm, Seth Myers <sjmy... at (no spam) syr.edu> wrote:
Quote: Simple question...
If I have a set of variables and perform a PCA to create from them
a few non-correlated variables, will the slope coefficients for
these principal components (in a linear model) be the same no matter
how (order,etc) they are entered into the model?
Yes, provided the model contains no non-PC predictors that are
correlated with the PCs.
Quote:
For instance, I would like it if after the following two functions were fit, B1 from the first and second function would be the same.
y = int + B1*PC1 + B2*PC2 + B3*PC3
y = int + B1*PC1
B1 should be the same in both models.
|
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Tue May 13, 2008 3:20 pm
|
|
