Modelling trend in break-point estimation: an assessment of the heat tolerance and temperature effects in four Italian cities

This paper deals with detecting trend in break-point estimation when some segmented relationship is met in each sub-sample of data. Based on iterative fitting of linear model, the procedure does not depend on the type of the response variable and always reaches the solution in deterministic model. Here the method is applied in mortality time series data to model the heat-tolerance with respect to local temperature in four Italian cities. In order to take into account possible different heterogeneity in each town, an Extend Quasi Likelihood approach is used, allowing joint modelling of mean and dispersion.