Bioinformatics 2006 22(19):2453; doi:10.1093/bioinformatics/btl435
Corrigendum
Bioinformatics (2006) 22(10), 1245–1250Avoiding model selection bias in small-sample genomic datasets
Daniel Berrar, Ian Bradbury and Werner Dubitzky
The authors would like to apologize for an error in the calculation of the true error rate 
based on Equation (1), which entailed too narrow intervals for this equation.
On page 1248, the confidence interval for based on Equation (1) is given by = 15.00 ± 0.08. Here, the center is given as percentage (i.e. 15%), whereas the limits are not. The correct approximate interval is 
0.15 ± 0.07499. As a consequence, the statements concerning the comparisons of the intervals based on the different equations need to be corrected:
For M = 100 and 
= 0.15, the widths of the intervals are 0.14998 for Equation (1), 0.13978 for Equation (2) and 0.13826 for Equation (3).
For M = 300 and = 0.15, the corresponding widths are 0.08415 for Equation (1), 0.08079 for Equation (2) and 0.08048 for Equation (3). The confidence interval based on Equation (1) as shown in Figure 1 is too narrow. The corrected figure is shown below.
|
Page 1247, second column, lines 19–20: ‘[...] an approximate (1 – 
)%-credibility interval [...]’ should be ‘[...] an approximate (1 – )-credibility interval [...]’.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||