Table 1.
Differences between the 3 methods.
Fig 1.
Shape of the beta and binomial-scaled distributions with standard deviation σ = .1 and means p = .5, p = .7, and p = .9.
Table 2.
Comparing distributional assumptions between the methods when σ = .1 and ρ = .1.
Table 3.
Comparing the Pezzoli, Hund, and Hedt designs for σ = .1, α = β = .1, and m = 10.
Table 4.
Comparing the Pezzoli, Hund, and Hedt designs for ρ = .1, α = β = .1, and m = 10.
Fig 2.
Plot of beta distributions with fixed mean coverages and varying standard deviation and/or intraclass correlation.
The dotted lines represent the mean coverages pl = .75 and pu = .9, which are the same in all 3 panels. Left: standard deviation is fixed for both distributions at σ = .1; middle: intraclass correlation is fixed at ρl = .05 (note: ρl = .12/(.75*.25)); right: intraclass correlation is fixed at ρu = .11 (note: ρu = .12/(.9*.1)).
Fig 3.
Top panel: Estimated σ and ρ as a function of for 37 areas in Minetti et. al [27], with loess smooth overlayed.
Bottom panel: Solid line represents OC curve for n = 60, d = 50 for pl = .75, pu = .9 design when σ and ρ are fixed at the mean value of (left) and
(right). Dashed line represents OC curve when σ and ρ vary over p according to the predicted loess smooth of
(left) and
(right).
Fig 4.
Top panel: Estimated σ and ρ as a function of for 20 areas in Greenland et. al [12], with loess smooth overlayed.
Bottom panel: Solid line represents OC curve for n = 60, d = 50 for pl = .75, pu = .9 design when σ and ρ are fixed at the mean value of (left) and
(right). Dashed line represents OC curve when σ and ρ vary over p according to the predicted loess smooth of
(left) and
(right).