Logistic Regression Model and Model Selection for Estimating the Risk of Natural Disaster

Abstract

We apply a logistic regression model to the data of snow damage occured in Oyabe city in 2004 in order to analyse the damage risk on sugi (Cryptomeria japonica) stand on the continuous basis. For specifying factors effecting the snow damage, we rely on the following three information criteria, 1) Akaike’s information criterion, 2) Baysian information criterion and 3) Bias-corrected Akaike’s information criterion for selecting appropriate ones. Our results are compared with those from the previous work with the use of the discrete regression tree model. Our experiments show that the ability to judge the degree of the snow damage risk is almost the same for both approaches according to a critical value by 50%. This implies that considering the amount of information on the risk as well as automatic specification of appropriate variables for the model, we can conclude that the proposed method is superior to the previous one.