Sample size estimate for a single proportion
Source:    Publish Time: 2013-04-06 20:25   1279 Views   Size:  16px  14px  12px
Author: Xuanqian Xie   Objective: To exam whether % of the correct classification of Test A is higher than 80%

Author: Xuanqian Xie

 

Objective: To exam whether % of the correct classification of Test A is higher than 80%.

 

P_L: the minimal accepted level, 80%.

P_E: the expected “% of correct”, 90%.

Side: =1

Alpha: =0.05

Method: the exact method for binomial proportion

Power: >= 0.8

Null hypotheses (H0): P_E = P_L

H1: P_E > P_L

Based on the above assumptions, a conservative the sample size is 94 to ensure the power greater than 0.8.

 

It should be noted that the statistical power does not monotonically increase with sample size, but has a “saw-toothed shape”. Although the minimal N can be 82 to reach power of 0.8, it is safe to use the conservative estimate, 94. The more details of the method I used can be found in Chernick & Liu 2002. (Chernick MR, Liu CY. The Saw-Toothed Behavior of Power versus Sample Size and Software Solutions: Single Binomial Proportion Using Exact Methods. The American Statistician, Vol. 56, No. 2. (2002), pp. 149-155).

 

 

 

 

Prev article:No prev product
Next article:No next product