Massachusetts case curves have been fixed

[CSSEGISandData]

This is in response to issues #3091, #3092, #3099, and #3116, which concerned the removal of antibody tests from probable cases in Massachusetts. We have successfully liaised with the state health department to receive corrected time series files for these locations. This issue will describe our procedure for fixing our time series files.

• For January 22-September 2: county level data is the confirmed case total by specimen date from the state of Massachusetts EXCEPT for Dukes and Nantucket county where we have relied on our historical time series as the data from the state is incomplete for privacy reasons.
• January 22-September 2: the unassigned, Massachusetts entry is sum of probable cases across the state + cumulative confirmed cases in the "unknown county" category.
• September 3 to present: Time series file has not changed as our method for accessing data is sound and this is the date where Massachusetts removed antibody tests.

The impact of all of these changes is that probable cases are always aggregated in the unassigned entry regardless of historical knowledge of their location. This is necessary due to the state of Massachusetts only reporting at the state level moving forward. Of note, the distribution received from the state of Massachusetts was for cases by specimen date rather than date of reporting, which has invariably shifted many cases back in time. We find this impact to be unavoidable.

Thank you all for your gracious patience. We are very happy to see this corrected.