I am NOT an epidemiologist, virologist, disease modeler, public health researcher, etc. These visualizations were made for my own curiosity and I thought they would be of general interest. It is possible that my data normalization is overly naive and skewed in some important way.
MD does not report total tests given by zip code and date (they do this at the county level). This means that if tests are doubled, you would expect cases to also double. This also means I cannot report the zip code level percent positive rate.
MD does not report (or I could not find) the methodology for how the zip code level data is recorded. It is possible/probable that cases that cannot be associated with a zip code are dropped. Therefore cases may be missing, which can potentially skew the data.
There is no information on WHEN the case was reported. As there are many steps in the chain to report a case (person feels ill - gets test - waits for result - result gets reported to MD - data gets added to the database) it is possible that trends you see may have actually occurred week(s) ago. I have seen a few modelers use a 14 day lag between infection -> positive test.
2020-08-07: My partner informs me "sexactly" is not a word. New news added.
2020-08-05: Another limitation added. Brief news section added.
2020-08-02: Limitations section added.
2020-08-01: Added county lines in white, ability to overlay zip codes labels when zooming in. Made legend more clear in MD chloropleth plot
Source code for this app (which also shows exactly where the data is from and the manipulations done in R) can be found