Pfizer wants to estimate the efficacy of its vaccine to prevent serious disease and runs the following regression on data across multiple jurisdictions.

%ICU = a + b*%VAC + e ,

where %ICU is the percentage of the population within each jurisdiction admitted to an intensive care unit with COVID-19, %VAC is the percentage of the population within each jurisdiction that is vaccinated, a and b are coefficients to be estimated, and e is an error term.

Pfizer data scientists find that the estimate for b is small and positive, and the t-stat value is equal to 3. This result suggests vaccination rates increase ICU admissions. Before recalling millions of vaccines, an intern from Sauder points out that, due to supply constraints, older people are far more likely to be vaccinated than younger people, and also more likely to be admitted to an ICU. Jurisdictions vary considerably in their population demographics.

The intern's suggestion is that the regression analysis is problematic because of

["reverse causality", "statistical insignificance", "functional form", "omitted variables bias"] .

The issue could be fixed by using a

["multivariate", "linear", "quadratic", "least squares"] regression specification.