Should account for the exponential increase in number of tests given

[philgoetz]

Re. the code in visualizing_growth_rates:

The entire US performed an exponentially-increasing number of tests throughout all of March.
This means every state will appear to show nearly exponential growth if you track the number of positives, even if the probability of any given person in that state having the virus stays constant.

I plotted this for the US-level data for March 3-27, and found that the number of tests done per day explained ~98% of the variance in the number of positives found on each day. The two curves (columns 'positive' and 'totalTestResultsIncrease', each over days) are visually identical except for the Y-axis labels.

I recommend dividing the number of new positives on each day by the number of new tests on each day. You will, however, not get such pretty graphs, because many states failed to report negative tests on many days; on these days that fraction spikes up to 1.