FINGER EXERCISES TO BE DISCUSSED MON 28 OCT 2002
------------------------------------------------

1. Show that the gradient of the log-likelihood (the score function)
   for logistic regression is 

            X^T[Y - mu]

    where X is the matrix of covariates, Y is the column vector of 0's
    and 1's, and mu is the column of means of Y under the model.

2. Show that the negative Hessian and the Fisher Information for
   logistic regression are both  

            X^T Sigma X

   where X is the matrix of covariates and Sigma is the diagonal
   matrix with entries mu_i(1=mu_i) down the main diagonal.

3. NONLINEAR REGRESSION PROBLEM

4. Obtain the scoring algorithm for robust regression, in the general
   case discussed in class.

   Work out the details when rho(t) = Huber's function.

5. ROBUST REGRESSION PROBLEM