1. Refer to the May 2010 Qualifying Exam, Part 2, Question 5 (http://www.stat.sc.edu/sites/stat.sc.edu/files/attachments/may2010pt2.pdf).  
a) How would you choose start values for the first method (the Michaelis-Menton model)?  Fit the model using either PROC GLIMMIX or PROC NLIN. 
b) Given (a), how would you choose start values for the additional parameter in the second method?  Fit the model using either PROC GLIMMIX or PROC NLIN.
c) What might be an appropriate approach to model selection?

2. Using PROC GENMOD (unless you want to figure out the Negative Binomial likelihood for PROC NLIN), fit four loglinear models to the horseshoe crab data using width as a covariate: Poisson, Zero-inflated Poisson, Negative Binomial and Zero-inflated Negative Binomial.  Compare the Poisson models by examing the difference in their respective values of -2logL (Use Full Loglikelihood); it should have a chi-squared distribution with one df.  Use the same approach for the Negative Binomial models as well.  Now compare all four by examining AIC; which seems to fit the best?