Extra Exercise 17
-----------------

(a) Supplementary Exercise 9.2: Prescription of tetracycline drugs.

One response variable (whether a doctor prescribed tetracycline drugs to
at least one patient under the age of 8, or not), and the three county
types form an explanatory varible with three categories. The statistical model 
is 3 independent binomial distributions (model I), one for each county type, 
and the assumptions correspond to 3 independent binomial settings.


(b) Supplementary Exercise 9.3: Alcohol and nicotine consumption for pregnant mothers.

Two response variables (alcohol consumption prior to pregnancy (none, 0.01-0.10, 
0.11-0.99, 1.0+) and nicotine intake during pregnancy (none, 1-5, 16+)) observed for
452 expectant mothers interviewed during their pregnancy. Despite that one of the 
variables refers to the period prior to pregnancy, there is no indication that the
group of mothers involved in the study were specifically selected on the basis 
of their pre-pregnancy alcohol consumption levels. Therefore it is fair
to say that both variables were unknown prior to data collection in the
interview. The study focuses on outcomes for their 4-yr old children, and 
naturally these variable were known by then, but that is not of relevance 
for modelling the relation between these two variables. The model is therefore 
a single multinomial on 4*3=12 categories - the 12 possible combinations of 
the categories of the two variables. This is the model II, or a model/design 
to test independence between alcohol and nicotine intake.


(c) Supplementary Exercise 9.22: Performance of a stuck fund.

Two response variables (the performances (winner/loser) of 240 stock funds
over two consecutive years). The design could also be termed two paired
samples because the same funds were observed twice. The interest is in
whether any (positive) association exists between the performances in
the two years, contrary to other situations with two paired samples
where agreement or comparison of proportions is of interest. The
statistical model is a single multinomial on 2*2=4 categories - model II
to test independence between the two variables, or in this case between
the performances in the two years.


(d) Supplementary Exercise 9.57: Treatment for cocaine addiction. 

One response variable (cocaine relapse no/yes), and three treatment groups
forming an explanatory variable. Note that the data table has been presented
with the explanatory variable along the rows, contrary to the examples
considered so far. So in this case the statistical model is a binomial
distribution for each row, and model I to compare independent
populations.