|
Sample Size
and Power
To effectively compute sample size,
we may require information about your study as well previous studies.
Below is an outline of the necessary information followed by a brief example.
-
Express research hypothesis
as a statistical (null) hypothesis
-
Determine whether alternative
hypothesis is one or two directional
-
Specify level of significance
(e.g., 0.05)
-
Specify estimate of standard
deviation of variables to be analyzed
-
Specify estimates of group
means (effect size)**
-
Specify desired power (e.g.,
80%) **
-
Determine sample size per
group **
Example
A clinical trial is conducted to
investigate the efficacy of treatment A versus treatment B for glucose
control. The primary outcome variable is A1C at 6 months. The statistical
hypotheses are given below.
| |
Null Hypothesis: Mean A1C the same for two treatment
patient “Populations” |
| |
Alternative: Means not the same |
Parameter estimates were based
on a small sample of diabetic patients given treatment A. Their mean A1C
is 8.4 with standard deviation 2.3. The investigator wants p-values equal
to or less than 0.05 to reject the null hypothesis and power at least
80% to reject if mean for B is 1 or more units different from 8.4.
The required sample size is
85 for each of the two treatment groups for 170 total patients.
|