Wednesday, April 29, 2020

Messy Data

When you are looking at data and trying to find a trend, you can be stymied by smaller trends within the longer trend and they can dwarf the trend you are looking for.

Case in point is a long term trend in global temperature. The variation one sees every day is larger that what one might see in thousands of years, same for seasons. And added to that is the need to take readings all over the planet.

When it comes to looking for trends in the US Covid pandemic, we have unique sorts of problems in data messiness. There are weekly anomalies--the weekend changes people's schedules and so death reporting may lag by different rates on different days. Here, I spotted a very clear-cut Sun, Mon, Tue pattern from the Worldmeter site:



You see hints of this throughout the month or so, but this week and last week are especially clear. Another great example is Sweden.