TY - JOUR TI - Bayesian model averaging with exponentiated least square loss DO - https://doi.org/doi:10.7282/T3348HD1 PY - 2013 AB - Given a finite family of functions, the goal of model averaging is to construct a procedure that mimics the function from this family that is the closest to an unknown regression function. More precisely, we consider a general regression model with fi xed design and measure the distance between functions by mean squared error (MSE) at the design points. In this thesis, we propose a new method Bayesian model averaging with exponentiated least square loss (BMAX) to solve the model averaging problem optimally in a minimax sense. KW - Statistics and Biostatistics KW - Bayesian statistical decision theory KW - Least squares KW - Regression analysis LA - eng ER -