Teaching Likelihood Inference and Inference to the Best Explanation with Zocchi 100-Sided Dice

Oregon Convention Center 

Likelihood methods have numerous applications in bioinformatics and underlie a variety of fundamental tools of the discipline. To familiarize students in bioinformatics (and other fields) with using such methods, we developed a one-period (75-minute) module. To facilitate active learning and student engagement, we incorporated a tactile element in which students use Zocchihedron dice to generate simulated data. ("Zocchi" dice have 100 approximately equally-likely sides.) We begin the module by defining inference to the best explanation and discussing everyday examples of using it. Next, we introduce likelihood as a rigorous means of comparing explanations or hypotheses. We then have students use Zocchi dice to simulate the formation of short DNA sequences assuming two possible distributions of the four component nucleotides (A, C, G, T). Finally, students calculate the likelihood ratio and use it to infer the nucleotide distribution. We finish the module by discussing various aspects of likelihood inference, e.g., the longer the DNA sequence, the more accurate the inference.