I am control-engineer, I achieve accurate speed-control of noisy-feedback, high-inertia mechanical systems.

The hospital-deaths data has a huge weekend noise-blip, so I apply a 7-day rolling average filter. This transforms the data at the expense of a 3.5 day average lag.

For me it is crazy to use raw-data without this filter, the process-lag is more than 2-weeks and the consistent week-cycle coloration is huge.

So I would 7-day rolling-average deaths and sewage data and look at the curves, which I think is roughly what original paper did.

I would then apply statistical tools to estimate lag and correlation.

To use statistical tools to estimate lag and correlation without some sort of 7-day filter seems to me to be a big mistake ! ]]>

It is nice to have someone interested in sociology and data science. My name is Pablo, I would like to know if you have any package in mind for sentiment analysis.

Best regards

]]>I remember the same thing, which happened here around 1969 (we followed the UK’s lead). My key memory is that the government issued every kid with a reflective armband (in those days everyone walked to school).

]]>