Improvements to DSL for ASKEM

[cthoyt]

• Propositional predicates
• Represent expectations
• Represent differences of probability distributions
• Implement counterfactual variables and values

[djinnome]

OK, this is currently the query for the December demo Question 1: Looking back to this time, which interventions could we have implemented to keep below a hospitalization threshold of 3k covid patients, over the winter 2020 season (Dec. 1st 2020 to March 1st 2021)? (Can this be stated probabilistically? How likely would an intervention have enabled us to reach our goal?) .Very limited social distancing and masking policies (say this would only apply to healthcare settings, assume 5% decrease from normal contact/transmission levels) beginning right at the start of the period on Dec 1, 2021, through March 1, 2021. Stronger social distancing and masking policies, wait until after holidays and begin on Jan 1st, 2021 through end of 3 months (until March 1, 2021). What is the severity of intervention required for this option to have been successful? (for CHIME model this intervention maps to % decrease from baseline transmission levels). Modeling constraint/goal: keep covid hospitalizations < threshold over next 0<=t<3 months (edited)

And this is the query for Question 2: What vaccination rate(s) would these two groups need to have over the next 3 months, in order to lower the observed case rate for those age groups below 10 cases per 100k population?

[djinnome]

Question 3a: Which of these models works better under training conditions? Which one should I trust more under certain conditions? Which one performed better during 𝑡 time period?

Question 3b: Which of these models works better under holdout conditions? Which one should I trust more under certain conditions? Which one performed better during 𝑡 time period?