Thursday, May 6, 2021

Trying to Predict the Future


Since the beginning of the pandemic, we've tried to follow the science, but the science seems to flip-flop every so often. We hear about this model or that model, but most people have no idea what that really means. Some at least understand that it is a computer simulation that seemingly predicts the future. Like the climate models. I came across an article on using a SEIR model (Susceptible | Exposed | Infectious | Recovered) to predict herd immunity. This led me to a Wikipedia entry about "Compartmental Models in Epidemiology" in which SEIR is but one of the models in use. If you're up on your ordinary differential equations, you'll find this fairly straightforward. Each person is in one compartment or another and the diff-e's for each are taken with respect to time (i.e., ~ / dt). You'll also find three assigned constants, "β", "γ" and "N", where "β" is the average number of contacts per person per time, "γ" is the infectious time period, and "N" is the total population being analyzed. "N" is easy to figure out, but the value of "β" is essentially a best guess based on the data at any given time.
The accuracy of the selection of a value for "β" affects the accuracy of the outcome of the model. E.g., if the time unit is a day, then "β" is the number of contacts each person has with someone else on a given day. How many is that? Five? Ten? One? It's only a guess and it may explain why so many models are at odds with each other.


No comments:

Post a Comment