EViews.com

Hi everyone,

I have a sample of 400 time series that I would like to explore using principal comoponent analysis. I can enter all 400 factor. However, only a maximum of 200 are used to construct the actual factors. Here my questions:

- What does the max number of factors entail? Are some of the original series excluded when building the factor components? (the number of Eigenvalues displayed seems to suggest that this is so)

- How does EViews decide which series to include when building the factor components?

- What does this mean with respect to the interpretation of the factor loadings?

Any help is greatly appreciated...
Version: EViews6

I'm not certain I understand what you are saying here. Do you have a workfile you can provide containing the group of series for which you wish to compute the PCA.

sorry about that. I found that the +400 variables I tried to group together were not all included. It seems as if only 200 of the original series are grouped together. Is this a default cut-off? How can I group more series together?

Shouldn't be. How did you create the group?

Group selection as follows:
1. Select all variables (except residuals and intercept)
2. Object/New Object
3. Select group with unspecified name
4. Name group
This works fine but does not appear to include all variables. However, via View/Group members and Copy/Paste I was able to add the missing series. Maybe not the way one is supposed to do it but worked for me. Cheers

It is interesting, but yes EViews does select the first 200 series, if you follow those steps. Try "right click" and "open as group" after you select all the series in the first step.

My guess is that you are running out of room in the dialog so it isn't getting all of the names. trubador's suggestion will always work since we will get the names from the selection.

Yip, works for me! Thanks for the input, guys! New to EViews so sorry about the somewhat basic question...