Analysis of vesicular-arbuscular mycorrhizal colonization data with a logistic regression model

Vesicular-arbuscular mycorrhizal (VAM) infection is usually expressed as percentage of root length colonized. The frequency distributions of the data are often non-normal and may follow a negative binomial distribution. Data transformation, such as an arcsin of percentage colonization? may be used to help colonization data satisfy the normal distribution assumption, but is not always successful. In this paper, we compare ANOVA and logistic regression model (LRM) analysis of data on the effect of phosphorus fertilization and corn cultivar on VAM colonization. Transformed data did not fit a normal distribution, and we propose the LRM as a better model for statistical analysis of VAM colonization. The LRM is more accurate because (1) this model assumes a binomial distribution, (2) it incorporates the original sample size into the probability estimation, and (3) the model uses non-transformed data, which are easier to interpret.