If you used the function \(g l m(y \sim r x 2\), family=Gamma(link= identity)) in the preceding exercise, you may have experienced a failure to converge. Write your own Newton-Raphson function with steps on the log scale, which forces the \(\beta\)-components to be strictly positive. As part of this exercise, you will need to compute the Fisher information matrix, \(\mathrm{t}\left(\mathrm{X} / \mathrm{mu}^{\wedge} 2ight) \div \% \mathrm{X}\) for \(\beta\). Report the value of \(I_{\beta}\) at the null hypothesis \(\hat{\beta}_{0}\) and also at \(\hat{\beta}\). What does this tell you about the Wald-Wilks discrepancy?