|
Abstract:
|
Finite mixture models are used when sampling from a population that consists of subpopulations with similar characteristics, but with at least one separate identifying mark. Here, a mixture of regressions model is considered to describe pairs of data (X, Y). For example, the mixture of regressions is used in modeling judgments for a child's thinking strategy when given certain tasks (Thomas and Horton 1997). The goal of this talk is to introduce allowing the mixing weights of the model to depend on the predictor. This enables the investigator, given the predictor, to determine which of the K regression equations in the mixture model best predicts the value of the response.
|
