home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.stat
- Path: sparky!uunet!mcsun!news.funet.fi!network.jyu.fi!vms
- From: vms@jyu.fi (Veli Matti Suppola)
- Subject: Robust correlation coefficient
- Message-ID: <1993Jan9.031500.24239@jyu.fi>
- Keywords: robust
- Organization: University of Jyvaskyla, Finland
- Date: Sat, 9 Jan 1993 03:15:00 GMT
- Lines: 32
-
- Hi!
-
- I'm interested in robust correlation coefficients.
-
- I will examine traditional correlation coefficients (Pearson,
- Spearman, Kendall). These correlation coefficients are
- sufficiently robust against a substantial number of outliers.
- That is, they do not have high breakdown points.
-
- I will try to find new correlation coefficients that have higher
- breakdown points. I have found two 'new' coefficient but the more
- the better. I will compare these coefficients through simulation
- and try to find the best.
-
- Robust correlation coefficients I have found:
- *Least Median of Squares (Abdullah 1990)
- *The Greatest Correlation Coefficient (Gideon 1987)
-
- I don't have enough material to examine different robust
- correlation coefficients. So, if You know some elementary
- books about this area, please let me know. All types of
- articles about this area are very welcome.
-
- The second problem is that there are no programs, algorithms
- or statistical packages that can create robust correlation
- coefficients. So I have to make these programs by myself. If
- You know any programs or algorithms about this area, please
- let me know.
-
- Veli-Matti Suppola vms@jyu.fi
-
-
-