Out of sample forecast of Principal Components

[afritsche]

I want to generate an out-of-sample forecast (and backcast) of a principal component I estimate over a specific period. (Effectively I want to generate an index using the principal component, but only change my base weights periodically, rather than re-estimating each time as new data becomes available). This means that I need to recreate the principal components based on the weights that EV calculates.
When I apply the loadings to my my underlying data to replicate the principal components that Eviews generates, I get something different from the PC that EV generates I'm only interested in PC1). I understand that Eviews scales/centers the data before obtaining the principal components. However, I don't know how it does that. As such, I can't reverse the process.
How do I go about doing that? Or, how do I find out what scaling/centering Eviews applies before it calculates the PC? - Thanks

[EViews Glenn]

It's the same operation that goes into computing correlations since that's the fundamental for pc analysis. Centering with respect to mean and dividing by the standard deviation (df or not df corrected as desired). Note that this because tricky if you allow for unbalanced computation.