Coronavirus Outbreak – It’s Time to Overlay the Moving Average

(Click on the image to enlarge)

Now that the coronavirus outbreak has been in full swing for over six weeks, it's a good practice to study the daily data points alongside a simple trendline representing the 3-day moving average. This smoothes out the daily reporting anomalies, especially the the weekends and holidays. As the liquidity increases, the 3-day average can be extended to a 5-day and 7-day averages. 

This is very similar to daily closing vs. weekly closing prices for stocks and indices. When an extended series is available, the weekly closings provide a smoothed out trend. 

Likewise, if the current outbreak continues to plague us for an extended period of time, researchers would be looking at weekly trends rather than the daily trends. When a particular short stretch has to be looked at more closely, say between 5/1 and 5/15, obviously the daily data points remain more critical.

The advantage of a moving average trendline is that it, unlike the linear regression line, does not suppress the seasonal peaks and troughs. For instance, the moving average trendline of an extended two year outbreak, as in the case of the SARS outbreak, will still show the bi-modal allergy seasons (e.g., Apr-May and Oct-Nov), thus allowing researchers the opportunity to address or flatten those seasonal peaks.    

The above graphic shows that while NYS peaked around 4/10, there were follow-up spikes on 4/15 and 4/25. The 3-day moving average trendline, however, has moderated them down with two lower low peaks. Overlaying the moving average trendline is useful for presentations as well. 

When a forecasting model is needed out of an extended time series, the Autoregressive Integrated Moving Average (ARIMA) model is very handy. Hopefully, we will not reach that point as far as this outbreak is concerned.

Stay safe!

Data Source: https://en.wikipedia.org/wiki/COVID-19_pandemic_in_New_York_(state)


-Sid Som
homequant@gmail.com