Dunnett's test

Dunnett's test is a statistical test developed by Charles W. Dunnett in the 1950s. It is used to compare multiple experimental groups with a single control group when the data are normally distributed and the variances are equal. Dunnett's test is particularly useful in the field of biostatistics and pharmacology, where researchers often need to evaluate the effects of various treatments against a control.

Overview

Dunnett's test is a post hoc analysis that is applied after an ANOVA has determined significant differences among group means. The primary purpose of Dunnett's test is to identify which means are significantly different from the control mean. This test adjusts the Type I error rate, maintaining it at a desired level, which is crucial when multiple comparisons are made, as it reduces the risk of identifying a difference as significant purely by chance.

Procedure

The procedure for Dunnett's test involves comparing the mean difference between each experimental group and the control group against a critical value from the Dunnett's distribution. This critical value depends on the number of groups, the total number of observations, the desired level of significance, and whether a one-tailed or two-tailed test is used. A significant result indicates that the mean of the experimental group is significantly different from the mean of the control group.

Applications

Dunnett's test is widely used in various scientific fields, including pharmacology, agriculture, and psychology, where researchers often test the efficacy of different treatments, drugs, or interventions against a standard or control condition. It is particularly favored when the comparison to a control group is more relevant than comparisons among all groups.

Advantages and Limitations

One of the main advantages of Dunnett's test is its ability to control the Type I error rate across multiple comparisons. However, it is limited to comparisons with a single control group and assumes that the data are normally distributed with equal variances among groups. When these assumptions are not met, alternative methods such as the Kruskal-Wallis test or Welch's ANOVA might be more appropriate.

See Also

• Analysis of Variance (ANOVA)
• Multiple comparisons
• Biostatistics
• Pharmacology