[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Interpreting PCA output

From   "Mona Mowafi" <>
To   <>
Subject   Re: st: Interpreting PCA output
Date   Thu, 24 Jul 2008 09:52:21 -0400

This is fantastic.. I will definitely take a look at your paper..Thank you!

Just to be clear on the point you make re: dummy variables with the
same factor; I follow F&P example in including multiple dummies with similar
factor b/c in any case, each dummy is acting as a separate component in 
the analysis and is only explaining away a percentage of the total remaining
variance, as I understand it.  So there is no overlap between them in the 
procedure.  I am not using a method for common variance (to identify factors)
as you may (?) have indicated below.. Just thought I'd clarify... 

Again, many thanks...

>>> "Stas Kolenikov" <> 7/24/2008 12:03 AM >>>
Oh, that Filmer and Pritchett stuff. Take a look at our working paper
Typically the first component is the measure of wealth, while the next
one or two are the measures of structure of that wealth (e.g.
urban/rural difference, etc.). But in your case, you have dummy
variables produces from the same factor (floor material, wall
material, etc.), and they are negatively correlated. Rather than
picking the common variation due to underlying wealth, the PCA now has
to work on explaining those extra correlations.

On Wed, Jul 23, 2008 at 1:55 PM, Mona Mowafi <> wrote:
> Dear Statalisters,
> I'm new to the list and hope you will be able to help me with an analysis problem before me.
> I have conducted a principal components analysis to identify principal components for 67
> underlying indicators or household asset.

Stas Kolenikov, also found at 
Small print: I use this email account for mailing lists only.
*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index