In the midst of the the COVID-19 epidemic, I was teaching a class on modeling and analyzing biomedical data at the University of Delaware. We switched the examples from HIV modeling to COVID pretty quickly, and I was happy with the outcome. This lecture is an example from near the end of the class that uses most of the techniques we covered. Using live infection counts from the state of Delaware coronavirus dashboard, we developed appropriate modeling hypothesis (including null hypotheses), developed a Poisson likelihood function for the measurement noise in population sampling, found maximum likelihood fits to the models, applied Akaike’s Information Criterion for model selection, and used Markov-Chain Monte Carlo methods to establish uncertainty bounds on the resulting fits. A roughly 2.5-fold decrease in the viral spread was seen on April 6, 2020, which corresponded well to the statewide school closure a week earlier.