In the context of Generalized Linear Models consider response variables Yl, which are assumed to be exponentially distributed. The specification for the model is EYC=exp(+x) (i) Specify the natural link function associated with linear predictor i=+xi [2] (ii) Given the maximum likelihood estimates ^=7.35 and s.e. (^)=2.13 construct an approximate 99% confidence interval for , and thus interpret the significance of this parameter. [3] (iii) Compare model A:EYi=, and model B:EYi=+xi, by testing whether =0, and thus identify which of the two models is more appropriate. To do this use the available measurements of scaled deviances: for model A 17.31, and model B 13.19. [5].