All models are wrong, but some are useful; Are we 'focused' while selecting this 'useful' model? 'Focused' model selection for dependent data

A statistical model typically can be defined as a collection of probability measures on some outcome space. All, except perhaps one of these measures are ‘wrong’. Traditionally, the selection of a model is done based on certain fit criteria (via some loss function such as least squares, maximum likelihood etc.), that is, the distance from the data to the model. Usually, these include AIC, AICC, BIC, DIC, HQC among others. ‘Focused Information Criterion’ is comparatively a new model selection approach, which is getting popularized nowadays. This approach essentially minimizes the MSE of the estimates of the parameter of interest (Focus parameter) using their asymptotic theory. In this talk, we will discuss some of the theory and applications of ‘Focused’ model selection for time dependent data.