Logistic regression is a useful tool for analyzing data that includes categorical response variables, such as tree survival, presence or absence of a species in quadrats, and presence of disease or damage to seedlings. The models work by fitting the probability of response to the proportions of responses observed. For instance, the number of outplanted seedlings in 50-tree rows that die from frost damage is an observed response. These observed numbers are converted to proportions which are then fitted by models that determine the probability that a seedling will die from frost damage. Normal distribution approximations to the proportions and the consequent analytical methods (e.g., regression and analysis of variance) can be used if large sample sizes exist for each experimental unit. However, logistic regression does not require large sample sizes for the data analysis to be feasible. Furthermore, it is possible to analyze individual tree data.
Download Working Paper 26 PDF file (878 KB)
To view this document you need the current version of
Adobe Acrobat Reader, available free from the
Adobe Web Site.
Updated November 04, 2009