Analyzing Data: What to do till the Statisticians Agree (Invited Presentation)

One of many skills expected of a scientist is the ability to analyze one's own data. Sometimes that is straightforward. Often, however, the data are complex, arising from unbalanced designs or—worse—undesigned observational studies. Further, statisticians themselves typically disagree about the best ways to analyze data. How should scientists address questions such as: Take a Bayesian or frequentist approach? What to do with data that are not independent? How large a sample size do I need? What about explanatory variables that are correlated among themselves? What to do about missing data? What if my variables are not Normally distributed? And, especially, Are significance tests OK or not? We will discuss questions such as these in a non-threatening manner and offer straightforward recommendations.