A0453
Title: Nonlinear least squares, model selection, and model averaging for social interaction models
Authors: Hon Ho Kwok - National Taiwan University (Taiwan) [presenting]
Abstract: A unified theory of nonlinear least squares, model selection, and model averaging is developed for social interaction models. First, although the generalized method of moments and maximum likelihood have been used to estimate social interaction models, nonlinear least squares consistently can also consistently estimate social interaction models. Since the least squares criterion functions coincide with Mallows' criterion, a model selection and averaging theory based on Mallows' criterion is provided. Second, an identification theory is developed based on the least squares criterion. Third, Mallows' criterion is proposed to be used for selecting networks, variables, and models.
