1 Statistics 601, Fall 2006 Assignment 7 On the course web page under Data is a link called “Simulated Additive Error Regression”, which you can use to obtain a file with the data shown to you in lab as a scatterplot. The first column of the data are values of a covariate (x) and the second column contains values of a response (y). These data were simulated from a model of the form, Yi = µi (β) + σ {µi (β)}θ i , (1) where µi (β) = β0 xβi 1 exp(−β2 xi ) i ∼ iid F (location/scale ) E(i ) = 0; var(i ) = 1 Using the model given in expression (1) and the given expectation function, estimate parameters of the model and produce interval estimates for the regression parameters (i.e., the parameters in µi (β)) and, if appropriate, for θ, using the following approaches, 1. Traditional generalized least squares for β with fixed θ and moment estimation of σ 2 . Clearly demonstrate how you arrived at your choice for θ. 2. The Pseudo-likelihood method of Carroll and Ruppert for β and θ and moment estimation of σ 2 . 3. Maximum likelihood assuming i ∼ iidN (0, 1) for β and σ 2 , with unscaled profile likelihood for θ (note that this should produce the mle of θ as well). 4. Maximum likelihood for all parameters including computation of inverse information assuming i ∼ iidN (0, 1).