ABSTRACT
The key idea of local modelling is explained in the context of least squares regression models. The simplicity, interpretability and its good statistical properties indicate that the local modelling approach can also be very useful in a wide array of statistical problems. Important developments in statistics in the past three decades include analyzing health sciences related data, in which the interest frequently centers on understanding how risk factors contribute to the survival times of a group of individuals. A family of useful models for data analyses is the class of generalized linear models. These models include many commonly used distributions such as the Gaussian, binomial, Poisson and gamma distributions, and can be used to analyze both discrete and continuous types of data. In many applications, the variance of the stochastic components can be large. This seriously affects the quality of the estimated functional relationship between the response variable and its associated explanatory variables.
