PCA - How to get back to the old data?
Posted: Thu Jul 10, 2014 12:47 am
Hi,
sorry for this stupid question, but i just can´t find out how to. I did a PCA in EViews on 3 time series, so as output, I get a 3x3 matrix of the Loadings/Eigenvectors. Now I want to reconstruct the original time series, so I need the scores. I let EViews calculate the 3 scores series, and as far
as I understood, multiplying the score with the loading matrix should return my original time series, right? However, it doesn´t seem to work.
Do I have to add the mean of the original time series? Which scaling do I have to use when calculating the scores? Did I forget anything?
Thank you for helping an absolute beginner!
Chris
sorry for this stupid question, but i just can´t find out how to. I did a PCA in EViews on 3 time series, so as output, I get a 3x3 matrix of the Loadings/Eigenvectors. Now I want to reconstruct the original time series, so I need the scores. I let EViews calculate the 3 scores series, and as far
as I understood, multiplying the score with the loading matrix should return my original time series, right? However, it doesn´t seem to work.
Do I have to add the mean of the original time series? Which scaling do I have to use when calculating the scores? Did I forget anything?
Thank you for helping an absolute beginner!
Chris