Minimum value thresholding and megaFIPS aggregation for low counts

[dshemetov]

Just wanted to track the conversation that started here. There is code that zeros out signals when the sample sizes are low due to privacy. It would be good to be clear about where these decisions are made and be uniform about them. A possibility is creating a separate step in the pipeline that provides this functionality. I have not looked at enough indicators to see where this occurs, so maybe it's not a big deal, but wanted to have a record of these concerns.

[krivard]

Related to cmu-delphi/delphi-epidata#195

[dshemetov]

Quidel has minimum sample size thresholding.