A multifaceted analytical approach for detecting effects on semen quality when using small sample sizes.

Driven by technical, logistical and economic limitations, detection of treatment effects on semen quality typically include the design and collection of small sample datasets. A consequence of these small sample studies is that they suffer low statistical power. Historically, researchers faced with small sample size studies have relied upon non-parametric analysis; however, this approach is still unlikely to tease out a true statistical significance based upon limited sample size. Here we propose a novel methodology that can be applied in small samples study situations that combines repeated measures ANOVA and Mixed-Effects linear regression models with Bayesian Linear regression modeling when evaluating for treatment effects on quantitative semen quality parameters. Using this methodology, we show that investigating the data with this multifaceted analytical technique results in improved reproducibility and sensitivity of the findings while minimizing the likelihood of Type 1 errors when combining the inference statistics from multiple models/methodologies using Bayes Factor analysis.