Premium
Correction
Author(s) -
Stallard N.,
Friede T.,
Posch M.,
Koenig F.,
Brannath W.
Publication year - 2010
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.3817
Subject(s) - sample size determination , statistics , sample (material) , range (aeronautics) , mathematics , optimal design , regret , power (physics) , computer science , algorithm , chemistry , materials science , physics , chromatography , quantum mechanics , composite material
We regret that there was an error in the computer program used to perform the calculations reported in ‘Optimal choice of the number of treatments to be included in a clinical trial’ by Stallard et al. ( Statist. Med. 2009; 28 :1321–1338). Although the general message of the paper does not change, we would like to present a correction of some of the numerical details. Corrected versions of Figures 1, 2 and 4 are given. Figure 3 is correct as given in the original paper, but shows the total sample size for the optimal design rather than the sample size per arm as stated in the legend. In the numerical example given in Section 4, when the two‐point prior distribution is considered, the optimal design controlling the assurance is that with 85 patients in each of two experimental groups plus a control group. The optimal design controlling the conditional power also includes both of the experimental treatment arms and the control, but with a sample size of 50 per arm. For the bivariate normal prior, the total sample size for the optimal design is 81, so that the required sample size per group is 27.1. Designs to minimize the total sample size with assurance (upper row) or prior conditional power (lower row) of 0.8 for a range of discrete prior distributions given by E (θ 1 ), E (θ 2 ) and ρ. Heavy lines border the area in which it is optimal to include two experimental treatments and light lines give contours of total required sample size.2. Designs to minimize the total sample size with prior conditional power of 0.8 for a range of discrete prior distributions given by q 01 and q 10 . Heavy lines border the area in which it is optimal to include two experimental treatments and light lines give contours of total required sample size.4. Designs to minimize the total sample size with prior conditional power to reject all false null hypotheses of 0.8 for a range of discrete prior distributions given by q 01 and q 10 . Heavy lines border the area in which it is optimal to include two experimental treatments and light lines give contours of total required sample size.We apologize for these errors and are grateful to Willi Maurer for drawing them to our attention.