On Thu, 1 Feb 2007 07:45:03 -0500, Susan Heath <[EMAIL PROTECTED]> wrote:
>Ecologgers, > >Has someone out there used breeding bird survey trend data as a >dependent variable in a multiple regression? If so can I discuss with >you whether you transformed this variable and if so what transformation >you used. Hello, Sorry for being a bit harsh, but you should not ask this question. Or let me say it different, the question cannot be answered by anyone without seeing it. If you apply regression analysis, the assumptions are homogeneity, independence, no patterns in your residuals, normality (at each X value), and fixed X. It all depends which one is violated; in some cases you may want to transform, in other scenarios you may want to continue with GLM, GAM or mixed modelling, or even GLMM or GAMM. There are examples of some of these methods in www.springer.com/0-387-45967-7 (publication date: 23-3-2007), and also in: Analysing Ecological Data using GLMM and GAMM with R. by: Zuur, AF, Ieno, EN, Smith, GM, Walker, N. (Expected publication date: 2008). Publisher: Springer. But that take a bit longer. The message is: provided you are willing to apply either mixed modelling, GLMM or GAMM, there is no need for a data transformation on your bird data. In many ecological examples I have seen, heterogeneity is part of the natural system, and is too important to remove it with a data transformation. As to the suggestion made by the first posting, I would not apply a transformation on a richness diversity index. Alain Zuur www.highstat.com > >Also, does anyone know how to get SAS to compute delta AIC values? I >can get it to calculate the AIC value for each case but not delta AICs >and calculating them by hand will really be a pain. Thanks for any info. > >Sue >-- >Susan A. Heath >George Mason University >Environmental Science Department Fairfax, VA >Secretary, Virginia Avian Records Committee >=========================================================================
