Much ado about something: response to Haughton's reply to Duvendack and Palmer-Jones

"Much ado about something: response to Haughton's reply to Duvendack and Palmer-Jones." Journal of Development Effectiveness, 3(2), pp. 302–303

Notes

1. At the more usual 5 per cent level, the result for the ‘relatively robust’ example in Abou-Ali et al. (2010) becomes sensitive at Γ = 1.05, which is an odds difference that is surely suggestive of a vulnerability to hidden bias.

2. Elsewhere Rosenbaum states of this example: ‘fairly small biases, too small to easily detect, could readily explain the observed association [between coffee and myocardial infarction]’ (2002, p. 129).