Question
In a classic study, Milgram et al . “lost” stamped envelopes with ? ctitious addresses (Medical Research Association, Personal Address, Friends of Communist Party, and
Friends of Nazi Party).* One hundred letters with each address were distributed among four locations (shops, cars, streets, and phone booths) in New Haven Connecticut,
with the following results:
ADDRESS RETURNED NOT RETURNED TOTAL
Medical Research Association 72 28 100
Personal Address 71 29 100
Friends of Communist Party 25 75 100
Friends of Nazi Party 25 75 100
Total 193 207 400
(a) Using the .05 level of signi? cance, test the null hypothesis that address does not matter in the underlying population.
(b) Specify the approximate p -value for this result.
(c) Assuming c2 is signi? cant, estimate the strength of this relationship.
(d) How might these results be reported in the literature?
(e) Collapse the original 4 x 2 table to a 2 x 2 table by combining the results for the two neutral addresses and for the two in? ammatory addresses. Calculate the
oddsratio for returned letters.