Using the beta-binomial distribution to characterize forest health

The beta-binomial distribution is suggested as a model for describing and analyzing the dichotomous data obtained from programs monitoring the health of forests in the United States. Maximum likelihood estimation of the parameters is given as well as asymptotic likelihood ratio tests. The procedure is illustrated with data on dogwood anthracnose infection (caused by Discula destructiva) in the southeastern United States. The parameter estimates have important biological interpretation, and tests of hypotheses are more meaningful than traditional statistical analyses. The value of a modeling approach to dichotomous data analysis is emphasized.