Testing Commonly Held COVID-19 Hypotheses
with Correlational Analyses
Please NOte: This website is in an early stage of construction. the goal is to generate a lively discussion of the findings, and to allow for continuing updating as new data and analyses are considered.
While correlation does not prove causation, lack of correlation is rather strong evidence that there is no causal relationship. This website uses American state population level data to test whether two hypotheses are supported by simple linear regression coefficients of determination (R-squared):
Were Non-Pharmaceutical Interventions (NPIs) effective in lowering pandemic death rates?
Was the SARS-CoV-2 (sars2) virus the principle cause of pandemic deaths?
ABSTRACT
We use correlational analysis of the fifty American states to test two hypotheses: (1) Non-Pharmaceutical Interventions (NPIs) were effective in reducing COVID-19 (covid) mortality, and (2) most of the deaths during the pandemic were caused by the SARS-CoV-2 (sars2) virus. If correct, then these hypotheses should be verifiable by correlating population level excess death rates with (1) government policies aimed at saving lives and (2) parameters shown at the individual patient level to predict higher risks of mortality, e.g., old age.
We measure pandemic mortality as the degree to which actual all-cause deaths exceeded the CDC’s projections of expected all-cause death rates for the three years 2019 through 2021. We apply confounder analysis to our more surprising simple, one-variable regression results by using multiple regressions to check for potential omitted variable bias.
Our statistical analyses suggest that NPI intensity had no discernable impact on excess death rates, that increased mask usage was associated with increased death rates, that states with older populations tended to have lower excess death rates, and that higher poverty rates were the most likely cause of higher excess death rates.
Based on these analyses, we conclude that the lack of expected correlations of excess death rates with measures of NPI actions is rather strong evidence that NPIs did not reduce pandemic mortality at the state level. We also find that the lack of state level correlations with factors thought by healthcare experts to influence covid mortality rates at the individual patient level is hard to explain if the sars2 virus was responsible for most of the excess deaths.