Online Program

Stats For Rats: Educating the Laboratory Scientist on Statistical Methods other than the T-test (302611)

 View Presentation

Keywords: education, statistics in lab science

Problem: One area of research that has a need for increased statistical education is basic lab science. Many laboratory scientists perform their own statistics; however, it is not always clear to them how to analyze their data. This is especially true in cases with multiple cell lines, multiple runs, and replicates. Method: Through a workshop series conducted through Emory Children’s Center (ECC) and Children’s Healthcare of Atlanta (CHOA), we educate bench scientists on experimental design and statistical methods. We discuss common statistical misconceptions and miss-practices, and introduce alternative methodology to the usual two-sample t-test for data analysis. Results: Our first motivating example includes one cell line with peripheral blood mononuclear cells with and without different reagents and viruses. Specific cytokines were measured at different time points to measure the effect of multiple factors. We educated the laboratory scientist why a t-test will give an incorrect conclusion and how we can detect interactions not achievable through univariate tests. We demonstrated that ignoring correlations between observations could lead to misleading results.