I have a large data set and want to conduct a multiple linear regression but
there is high mullicollinearity among my explanatory variables. In order to
adjust for this I have conducted a principal components analysis and then
used the orthogonal components as independent variables in the regression
equation. I have back-calculated regression coefficients for the original
variables by multiplying the eigenvector matrix by the vector of
component regression coefficients (after removing components that werent
significantly related). So, my question is, can I calculate significance
levels, confidence intervals or standard errors to my back calculated
coefficients?...and if so how??
Thanks, Claire
Claire Wells
School of Biological Sciences
Royal Holloway
Egham
Surrey
TW20 0EX
Email - [log in to unmask]
|