Multiple comparisons in the randomization analysis of designed experiments with growth curve responses

Area of Science:

• Biostatistics
• Experimental Design
• Animal Science

Background:

• Comparing multiple growth curves in randomized experiments often involves complex statistical assumptions.
• Controlling the overall Type I error rate (alpha) for multiple comparisons is crucial for reliable results.
• Existing methods may not be flexible or free from standard assumptions.

Purpose of the Study:

• To develop a randomization approach for multiple comparisons of growth curves in randomized experiments.
• To enable prespecification and evaluation of exact Type I error rates for these comparisons.
• To offer a procedure free from standard assumptions in growth curve analysis and multiple comparisons.

Main Methods:

• Developed a novel randomization procedure for multiple comparisons of growth curves.
• The method allows for prespecification of the overall Type I error rate.
• Evaluated the Type I error probability for each component test within the procedure.

Main Results:

• The proposed randomization approach provides exact control over the Type I error rate.
• The procedure is applicable to various experimental designs, including Youden square designs.
• Demonstrated application in comparing mean growth curves of hormone levels across four treatments in an animal experiment.

Conclusions:

• The developed randomization method offers a robust and assumption-free approach for multiple comparisons of growth curves.
• This technique enhances the reliability of findings in growth curve analysis, particularly in animal studies.
• The procedure facilitates accurate pairwise comparisons among treatment groups.