22S:152  Homework 7 
      Fall 2002
      Due Fri. 11/01 in class


       Consider the data on counts of snow geese flocks that was
       presented in class.  The dataset is "snowgees.dat."
       You will be doing for observer 2 a similar analysis to what
       we did in class for observer 1.

       a. Produce a scatter plot of photo versus obs1.  Does it
          suggest that a simple linear regression model might be 
          appropriate? Why or why not?
 
       b. Why is it sensible to regress photo on obs2 rather than
          obs2 on photo?

       c.  Compute the regression of photo on obs2 via ordinary
           least squares and test the hypothesis that beta1 = 0
           at the alpha = .05 significance level.  State the
           meaning of this hypothesis, and the result of the test.
           Also produce a plot of the residuals versus the predictor
           variable.

       d.  Repeat part c, except for the regression of sqrt(photo)
           on sqrt(obs2).

       e.  Repeat part c, except assume that the var(e_i) = x_i^2 sigma^2.
           Refer to section 6.6 of the Chatterjee, Hadi, and Price
           textbook (photocopy attached)  for how to do this.
           Check your residuals vs. 1/obs2 plot.  Does it look better
           or worse than the residuals vs. predicted values plots in
           parts c. and d.?

       f.  Which analysis (c, d, or e) seems to satisfy the assumptions
           of linear regression the best?