Research Design, Biostatistics, and Literature Evaluation
Two-tailed tests distribute half of the ฮฑ value at the far left hand and far right hand ends of the
distribution
ii.
One-tailed tests distribute all of the ฮฑ value at the far left hand or far right hand ends of the
distribution
Type II error (beta [ฮฒ] error): Fail to reject the H0 when, in fact, there is a difference between groups
(i.e., the H1 is true). Decisional threshold to set ฮฒ at 0.20.
| d. | Power (1-ฮฒ): The ability to detect a difference if one truly exists. Contingent on sample size. However, |
|---|
this is an estimate and may be inaccurate (usually based on previous literature).
Sample size is the number of subjects or samples collected or assessed within a study and is
influenced by the following:
| (a) | Treatment effect, or difference within the predicted outcomes that is considered to be |
|---|
clinically significant (i.e., difference in control vs. treatment group mortality rate)
| (b) | The variability in the predicted outcomes (i.e., the variability in baseline mortality) |
|---|---|
| (c) | What is deemed to be an acceptable rate of type I error (ฮฑ) |
| (d) | Type II error rate (ฮฒ) |
ii.
Typically the greater the effect and the lower the variability, the smaller the sample size required
iii.
Decreasing the rate of type I error (ฮฑ) and/or increasing the power (decreasing ฮฒ) increases the
necessary sample size.
Continuous
Counting, chronological order (e.g., 7.0, 7.1, 7.2, 7.3)
Sample tests: t-test, Wilcoxon rank sum/Mann-Whitney U
Nominal
Categories or groups to which data are designated (e.g., race, sex, hypertension, heart failure)
ii.
Sample tests: Chi-square, Fisher exact
Ordinal
Categorical data that include an order (e.g., Likert scale)
ii.
Sample tests: Wilcoxon signed rank test, Wilcoxon rank sum test, Mann-Whitney U test,
Kruskal-Wallis 1-way test
Variable
Two Samples
>Two Samples
Independent
Related
Independent
Related
Nominal
Chi-square or Fisher
exact
McNemar test
Chi-square
Cochran Q
Ordinal
Wilcoxon rank sum
Mann-Whitney U
Wilcoxon signed rank
Kruskal-Wallis
(MCP)
Friedman ANOVA
Continuous with
no covariates
Student t-test
Unequal variance t-test
Paired t-test
1-way ANOVA
(MCP)
Repeated measures
ANOVA
Continuous with
one covariate
ANCOVA
2-way repeated
measures
ANOVA
2-way ANOVA
(MCP)
2-way repeated
measures
ANOVA
ANCOVA = analysis of covariance; ANOVA = analysis of variance; MCP = multiple comparison procedures.