A reversed-hazard-based nonlinear model for one-way classification

This paper proposes a new statistical method for one-way classification which is based on the (reversed) hazard function of the response variable. The model parameters are estimated according the maximum likelihood approach. Several testing procedures, e.g., generalized likelihood ratio test, are investigated to assess homogeneity of populations. A non-parametric method for data analysis is also proposed. Two data sets are studied using the obtained results.