• In this course almost all of our causal inferences are based on a finite-sample of data, and potentially extensions to a population of which the sample is representative

• Another kind of model-based imputation relies on mathematical/theoretical models to impute missing potential outcomes

• This can forecast future potential outcomes

• I’m going to look at this as it relates to the coronavirus

• While I’ll attempt to look at it from a causal perspective, I am NOT an expert in the spread of infectious disease

• I am sharing the results of models fit by others as an example of model-based prediction of potential outcomes… I am not in a position to verify or judge the validity of the models

• All of the models (all models!!!) are based on assumptions that may or may not hold (and which are outside the scope of this course).

• The future of this COVID-19 pandemic contains a huge amount of uncertainty. Models can attempt to predict the future in the face of uncertainty, but all come with limitations.

• Models Of Epidemic Predict Huge U.S. Death Toll; Scientists Hope For Better Outcome, npr.org, 3/31/20

• Based on: https://covid19.healthdata.org/projections

• https://www.wired.com/story/the-mathematics-of-predicting-the-course-of-the-coronavirus/

• Most based on the SIR differential equation model: S = susceptible, I = infected, R = recovered (or SEIR model, E = exposed)

• How is this causal?

• How might these potential outcomes change under different possible actions?

• We’ll focus on the causal effect of social distancing on the coronavirus death toll

flowingdata.com

Imperial College Report

Imperial College Report

• The following charts come from the two medium articles by Tomas Pueyo:

  1. Coronavirus: Why you Must Act Now

  2. Coronavirus: The Hammer and the Dance

• I like them because they are backed by data, but I have not personally verified the analyses or conclusions.

• Caveat: the author is not a statistician or an epidemioligist, and the article has not been peer-reviewed. Be careful taking things at face value!

• However, they have been endorsed by experts

https://medium.com/@tomaspueyo/coronavirus-act-today-or-people-will-die-f4d3d9cd99ca

https://medium.com/@tomaspueyo/coronavirus-act-today-or-people-will-die-f4d3d9cd99ca

https://medium.com/@tomaspueyo/coronavirus-the-hammer-and-the-dance-be9337092b56

• All of these predictions are based on many assumptions!

• This site is fantastic because the user can alter the assumed value of any of the model parameters, and see the resulting impact on the forecasts

• It also takes an explicitly causal perspective by modeling the effect of an intervention (social distancing) as a function of timing and how much we can limit interaction (transmission)

http://gabgoh.github.io/COVID/index.html

• The biggest thing we can do to thwart the coronavirus is reduce the tranmission rate, NOW!

• The transmission rate, R, is the number of people each infected person then infects

• \(R\) is naturally around 2-3. For the virus to die off, we have to get it below \(1\)

• How?

  • Social distancing!
  • Closing of schools, businesses, etc.
  • Shelter-in-place orders
  • Etc.