Principal Component Analysis-reconciliation
Posted: Thu Nov 21, 2013 4:38 am
I think there is a bug in PCA calculation methodology. Multiplying loadings with indicator series doesnt give me the PC1 vector as shown in the excel file.
Before I take the time to work through the calculations, I just want to double check that if you are doing verification of components computed from a correlation matrix you'll need to scale and center your data appropriately.
http://forums.eviews.com/viewtopic.php? ... nts#p20975
Beyond that, you'll have to provide additional information as it's not entirely clear from the spreadsheet calculations what you are doing.