Medical Biostatistics, Third Edition by Abhaya Indrayan

Author:Abhaya Indrayan [Indrayan, Abhaya]
Language: eng
Format: epub
ISBN: 978-1-4665-1390-7
Publisher: CRC Press LLC
Published: 2013-03-05T16:00:00+00:00

12.6.2.5 Power Analysis

Good researchers determine sample size on the basis of power required to detect a specified minimum difference and try to do study on the size so determined. Because of some exigencies, sometimes it becomes difficult to study as many subjects. This reduces the power. Sometimes, a study is done on specific n without basing it on power. The formulas given in Table 12.8 can be used to inversely calculate power for a specific n, when other pieces of information are available. This is illustrated in Example 12.17. Nomograms and other tables can also be used. This process, sometimes referred to as power analysis, is becoming a requirement these days as awareness has increased. In case you are doing a study yourself or consulting other studies to update your knowledge, examine if the power of the study is adequate to inspire confidence.

There is a considerable overlap in the literature about the term power analysis. It is used for calculation of power for given n as well as for calculation of n for specified power. Both calculations require prior specifications based on literature or previous experience. Real utility of power analysis is in designing a study and not so much in interpreting nonsignificance of the result once obtained. Nonetheless, midcourse power calculation can be done on the basis of the observed data and can tell you how much more n you need. Post hoc calculation can tell you what might have gone wrong. Use this experience for planning a better study next time.

There is another exception, though. If the study throws up very different p or SD than assumed earlier for calculation of sample size, and the effect size is found statistically not significant, recalculation of power with new p or SD is justified and can give you better leads about the strength of your result. This can lead to sample size reestimation as described in the next section.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.