In the demos we find NC Deer Crash statistics. NC_Deer_Crash.txt
The scenario is that you work for an auto insurance company whose 
management wants to alert people to the yearly increase in crashes
related to deer that occur at about the same time each year (mating 
season). 

(1) Plot the data and find the month of maximum deer related crashes. 

(2) Fit a model with that month (every year) as a dummy variable - it 
will have a 1 and 11 0s every year so lots of 1s. Include a (linear) 
time trend as well.  Give the t test for H0: no time trend. 

(3) Add a one denominator lag transfer function (the exp decay option)
to your model. Compare the fit in whatever way you like to that in part
2. Leave the trend in for now. What is the decay rate? Based on this 
rate, the accident count 2 months after the peak is estimated to be
what proportion of the peak? This is called a "dynamic regressor" in
our forecasting system. 

(4) It looked as though the month before the max is a bit elevated. 
Add a point intervention for that month (again, this is a twist - there
will be a 1 in this column every year). Compare the fit using whatever 
criterion you used above. Also put a 95% confidence interval on this 
pre-peak increase.  Is 0 in the interval? 

(5) Test for the linear trend in the part 4 model. Is there a 
significant increase or decrease?  Are the other terms significant?
Using your final model, add (at least) a year of months,compute and 
graph the forecasts and forecast intervals. 

(6) Give an executive summary that might help your insurance company 
management find the best timing for including an expensive "watch out 
for deer" reminder in their client mailings. Be clear and concise but 
complete.