Had some fun with numbers this afternoon. Copied and pasted case and death information by State from
https://www.worldometers.info/coronavir ... e_vignette (numbers as of end of yesterday), state by state vaccination rates from
https://usafacts.org/visualizations/cov ... er-states/, State by State percentage of population >=65 years old from
https://www.prb.org/resources/which-us- ... he-oldest/, and State by State population densities from
https://worldpopulationreview.com/state ... -densities pasted them into Excel.
All statistical tests were one tailed based on expecting that case and death rates would be negatively associated with vaccination rates and positively associated with percentages of populations >=65 and population densities.
Here is the correlation matrix:
"%FV" is percent fully vaccinated. Yellow highlights means the coefficient is significant at
>95% confidence. The 0.905215949 coefficient for percent fully vaccinated and percent boosted is important because it tells you that, as variables, both tell you close to the same thing. Most of you probably know that the absolute value of a correlation coefficient ranges from to 0 to 1 with 0 being no correlation and 1 being perfect correlation. 0.9 is a very high correlation. In subsequent regression analyses, I used percent boosted. I consider it to be the best indicator of relative tendencies of people in States to follow vaccine recommendations over time.
I did regression analyses to consider impact of percent boosted, %>=65, and population density on case and death rates at the same time. When I did it with case rates, %>=65 dropped out and I ended up with this Excel linest function output:
The fact that population density (Pop/mi2) is a "significant" factor in the model tells us that, even though population density was not "significant" when considered alone, it is "significant" when one "controls" for percent of population boosted.
When I did it with death rates, I got this:
What that says is that, when one "controls" for percent of population boosted, both percent of population >=65 and population density become "significant"
What all this tells you is that the "effect" of percent Boosted "overcomes" the effects of percent of population >=65 and population density on death rates and the effect of population density on case rates. Both percent population >=65 and population density have effects on death rates and population density has an effect on case rates. But you cannot "see" those effects when you just look at either percent population >=65 or population density alone because the percent vaccinated effect is strong enough in the opposite direction to "overcome" them.