QUESTION 9.5

In a randomized, placebo-controlled double blind study (Njagi, 2003), investigators collected data from pregnant women who were at high risk of malaria.

They were randomized to two groups: ITN+SP (insecticide-treated nets + sulfadoxine pyrimethamine, an oral malaria preventative medication), SP (oral malaria preventative medication only).

One outcome of interest was percentage of women who were anemic at delivery.

a) Is there evidence that the percent anemic is different in the ITN+SP group compared to the SP-only group? In other words, does the addition of the insecticide-treated nets to sulfadoxine pyrimethamine give additional benefit?

Conduct a test for the difference in proportions anemic equal to zero in the ITN+SP group vs. the SP-only group (including all steps and computing a 95% CI).

Group	P-hat:Anemic	X:Number Anemic	n
ITN +SP	0.286	30	105
SP	0.309	30	97

b) Can we conclude there is no difference in proportion anemic in the two populations? Explain briefly.

c) Compute and interpret the RR of anemia for the SP group vs. the ITN+SP group.

d) Briefly describe which measure we might select to summarize the data, RR vs. RD. In other words, what do the RR and RD communicate to the reader?

e) Curiously, suppose another studya much bigger studyhad the same estimates of anemia in the two groups.

Group	P-hat:Anemic	X:Number Anemic	n
ITN +SP	0.286	300	1050
SP	0.309	300	970

(Without doing another analysis . . .) Would the RD estimate be the same as above? Would the p-value for the RD be the same as above? Explain briefly why this shows why "contradictory findings" often occur in the literature.

f) We are given the summary data for this problem.Describe how the "observation-level" data set would be arranged for this problem. How many observations (rows) would the dataset have? In other words, what might var1 and var2 be named? List several "example rows" of the observation-level dataset.

ID	Var 2	Var 3