|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: RE: The dependent variable is a multi-proportion in actual values
From |
"Verkuilen, Jay" <[email protected]> |
To |
<[email protected]> |
Subject |
st: RE: The dependent variable is a multi-proportion in actual values |
Date |
Mon, 15 Jun 2009 13:09:29 -0400 |
Others have provided some answers. I cited John Aitchison's 2003
Compositional Data Analysis book already:
http://www.amazon.com/Statistical-Analysis-Compositional-Data/dp/1930665
784/ref=sr_1_1?ie=UTF8&qid=1245085601&sr=8-1
This puts compositional data analysis into a multivariate normal
framework, which overcomes the problems of the covariance structure of
the Dirichlet but buys other problems (sensitivity to outliers being the
worst).
An alternative is to convert the data to ranks and then use a rank
model, e.g., -rologit-. It depends on how much boost you get from the
metric information in the percentages. Often this seems to be "not
much."
JV
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/