Why is collinearity in ordination a problem?
Date: Fri, 20 Nov 2009 12:56:28 -0000 From: "Victoria June Ord" <v.j.ord at newcastle.ac.uk> Subject: [R-sig-eco] Why is collinearity in ordination a problem? To: <r-sig-ecology at r-project.org> Message-ID: <86406CC946606A4F91EAA53CF17C14CB011D78A8 at quarrel.campus.ncl.ac.uk> Content-Type: text/plain; charset="iso-8859-1" Dear All Looking at examples from Alain Zuur's 'Analysing Ecological Data'- it is apparent that collinearity is considered an issue in ordination. Could someone explain why that is?
It depends what you do. Collinearity means that covariates are related to each other; e.g. length and weight...or distance to the coast and depth (in most systems). These are all 2-way relationships....but it can also be at a higher dimension. And yes..it can also involve continuous and categorical (e.g. location in an estuary and salinity). It is perfectly legitimate to apply for example PCA on a set of highly collinear covariates. But you are in trouble if you apply one of these constraint ordination techniques, apply model selection techniques and come up with p-values. For an example of collinearity in regression, see: http://methodsblog.wordpress.com/2009/11/13/first-paper-now-online/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+wordpress%2Fmethodsblog+(methods.blog) In GLMs collinearity enters itself in a slightly different way due to the underlying maths. And the same holds for RDA/CCA type techniques. In my opinion collinearity is one of the most challenging problems with data analyses. Alain Dr. Alain F. Zuur First author of: 1. Analysing Ecological Data (2007). Zuur, AF, Ieno, EN and Smith, GM. Springer. 680 p. URL: www.springer.com/0-387-45967-7 2. Mixed effects models and extensions in ecology with R. (2009). Zuur, AF, Ieno, EN, Walker, N, Saveliev, AA, and Smith, GM. Springer. http://www.springer.com/life+sci/ecology/book/978-0-387-87457-9 3. A Beginner's Guide to R (2009). Zuur, AF, Ieno, EN, Meesters, EHWG. Springer http://www.springer.com/statistics/computational/book/978-0-387-93836-3 Other books: http://www.highstat.com/books.htm Statistical consultancy, courses, data analysis and software Highland Statistics Ltd. 6 Laverock road UK - AB41 6FN Newburgh Tel: 0044 1358 788177 Email: highstat at highstat.com URL: www.highstat.com URL: www.brodgar.com