Rolling principal component -- standardized variables
Posted: Wed Sep 19, 2018 8:41 am
Hi,
I have coded a rolling principal component analysis, and want to check to make sure that each series used in the PCA is standardized appropriately in the rolling windows (i.e. using each window's sample mean and standard deviation). Looking at the Eviews guide and in the forum, it seems like this type of normalization should be the default, but I cannot replicate the results.
I have two sets of data that I'm running through the code to cross-check results:
1) the raw data from the original series
2) the rolling z-scores of this raw data
The principal component series from these two sets of data should be the same, but there are some level discrepancies. Is there a way to check how the raw data is being standardized in each of the rolling windows as the principal component is being calculated?
I have Eviews 9.
Thanks!
I have coded a rolling principal component analysis, and want to check to make sure that each series used in the PCA is standardized appropriately in the rolling windows (i.e. using each window's sample mean and standard deviation). Looking at the Eviews guide and in the forum, it seems like this type of normalization should be the default, but I cannot replicate the results.
I have two sets of data that I'm running through the code to cross-check results:
1) the raw data from the original series
2) the rolling z-scores of this raw data
The principal component series from these two sets of data should be the same, but there are some level discrepancies. Is there a way to check how the raw data is being standardized in each of the rolling windows as the principal component is being calculated?
I have Eviews 9.
Thanks!