Ann Thorac Surg 2007;83:724
© 2007 The Society of Thoracic Surgeons
Correspondence
Type II Statistical Error
Gerald S. Weinstein, MD
New Mexico Cardiovascular Associates, 5455 Remington Rd, Las Cruces, NM 88011 New Mexico
(Email: gswmd{at}aol.com).
To the Editor:
Lingaas and colleagues [1] made a fundamental error in their conclusion that "off-pump coronary surgery provided the same angiographic graft patency as the on-pump technique." They fell victim to the common mistake of equating the lack of a statistically significant difference with lack of a (real) difference. There was, in fact, an observed difference in patency, ie, 80% in the off-pump group versus 87% in the on-pump group.
The authors failed to consider the probability of a type II (ß) error, which is defined as "the chance of erroneously failing to reject a null hypothesis that is, in fact, false" [2]. In common terminology, a type II error occurs when one fails to find a statistically significant difference, given that one exists. The complement of the type II error (1-ß) is the power of the test. The most common cause of the type II error is small sample size, leading to an "underpowered" study. Even though the authors recognized the "relatively low statistical power," they failed to do the appropriate power calculations.
In scientific literature, authors should refrain from drawing conclusions from a lack of a statistically significant difference unless they have done the appropriate power calculations and included them in the text of the article.
 |
References
|
|---|
- Lingaas PS, Hol PK, Lundblad R, et al. Clinical and radiological outcome of off-pump coronary surgery at twelve months follow-up: a prospective randomized trial Ann Thorac Surg 2006;81:2089-2095.[Abstract/Free Full Text]
- Colton T. Statistics in Medicine. Boston: Little Brown and Company; 1974. pp. 120-123.
Top
References