Multiple comparison procedures for simple one-way ANOVA with dependent data.

Area of Science:

• Statistics
• Experimental Design

Background:

• The assumption of independence is critical for valid statistical inference in experimental designs.
• Investigators frequently overlook or fail to verify the independence assumption in their analyses.

Purpose of the Study:

• To investigate the impact of violated independence on Type I and Type II error probabilities.
• To evaluate the performance of six multiple comparison procedures under varying degrees of data dependence.

Main Methods:

• A Monte Carlo simulation experiment was conducted.
• Simulations incorporated different patterns of means, hypothesis types, and levels of observation dependence.
• The study examined six distinct multiple comparison procedures.

Main Results:

• Violation of the independence assumption led to a failure in controlling the error rate per comparison for all tested procedures.
• Increased correlation (dependence) among observations positively correlated with an increase in per-comparison power.

Conclusions:

• The independence assumption is crucial for maintaining the integrity of multiple comparison procedures.
• Researchers must be cautious when independence is questionable, as it can lead to inflated Type I errors and altered power.