Power analysis of statistical methods for comparing treatment differences from limiting dilution assays

Area of Science:

• Biostatistics
• Cell Biology
• Immunology

Background:

• Limiting dilution assays (LDA) are crucial for quantifying cell populations.
• Accurate statistical comparison of LDA data is essential for reliable results.
• Existing methods for analyzing LDA data vary in their statistical rigor.

Purpose of the Study:

• To evaluate and compare the performance of six statistical methods for analyzing limiting dilution assays.
• To identify methods with appropriate Type I and Type II error rates for LDA data.
• To provide guidance on selecting robust statistical approaches for LDA.

Main Methods:

• A power analysis was conducted using simulated data from 1,000 independent replications.
• Simulated data mimicked a typical LDA with 12 dilutions and 24 cultures per dilution.
• Six distinct statistical comparison methods were applied to both simulated and real LDA datasets.

Main Results:

• Significant differences in Type I error rates were observed across the six methods.
• Likelihood ratio and Taswell's weighted mean methods demonstrated Type I error rates close to the nominal 5% level.
• Likelihood ratio and Taswell's minimum Chi-square methods showed superior power (lower Type II error rates).

Conclusions:

• The likelihood ratio and Taswell's minimum Chi-square methods are recommended for their statistical accuracy and power in LDA.
• Taswell's weighted mean test offers acceptable error rates for LDA analysis.
• The regression method is deemed unsuitable for scientific applications involving limiting dilution assays.