Skip to main content

Table 2 Effect of percent ascertainment on the Type I error rate

From: Reducing decision errors in the paired comparison of the diagnostic accuracy of screening tests with Gaussian outcomes

Paired screening

Disease

Percent ascertainment

True

Observed

Corrected

trial bias

prevalence

(Test 1/Test 2)

   
 

0.01

15/50

0.01

0.89

0.36

 

0.01

15/80

0.02

0.95

0.25

 

0.01

50/80

0.01

0.23

0.12

 

0.14

15/50

0.02

1.00

0.82

Yes

0.14

15/80

0.02

1.00

0.60

 

0.14

50/80

0.02

1.00

0.20

 

0.24

15/50

0.02

1.00

0.95

 

0.24

15/80

0.02

1.00

0.91

 

0.24

50/80

0.02

1.00

0.40

 

0.01

15/15

0.01

0.02

0.23

 

0.01

50/50

0.01

0.02

0.12

 

0.01

80/80

0.02

0.02

0.18

 

0.14

15/15

0.02

0.02

0.26

No

0.14

50/50

0.02

0.02

0.14

 

0.14

80/80

0.02

0.02

0.03

 

0.24

15/15

0.02

0.02

0.26

 

0.24

50/50

0.02

0.02

0.14

 

0.24

80/80

0.02

0.02

0.04

  1. Type I error rates are calculated over 10,000 realizations of the data for the hypothesis test of a difference in the full areas under the curves. The nominal Type I error is fixed at 0.05.