Across all disciplines, data analysis is vital for discovery and validation of scientific hypotheses and theories. But no matter how sophisticated the analysis may be, poorly collected data will significantly reduce the strength of one’s conclusions. An analysis of experimental data often provides strong evidence for causality compared with those based on observational data since the researcher is able to exert control over important factors associated with the outcome of interest. A well-designed experiment gives the researcher an increased chance for a successful study since major sources of variation can be accounted for that may otherwise mask differences between treatments.
This short course introduces the fundamental principles of experimental design and their application to three straightforward design scenarios. To motivate and promote these concepts, I use data from a simple, real-life experiment in which film canisters were launched based on three different amounts of water and Alka-Seltzer each. For each design scenario, a real-time analysis is performed in JMP using data from the film canister experiment, including plots, analysis of variance (ANOVA), and post-hoc analysis of treatment comparisons. Instructions will be provided to replicate the film canister analysis in JMP, which may be acquired by VT students at Student Software in 3240 Torgersen Hall 8 AM-5 PM Monday through Friday for a nominal fee. This course is most applicable to those in the physical sciences (e.g. engineering, chemistry, physics, etc.) and assumes the audience is familiar with ANOVA and post-hoc analysis.