 By now, we are quite familiar with the COVID-19 data. We noticed the data is reported per country, with dates in columns, and with additional latitude and longitude information. We have already plotted the data on a map in the previous video, and now we'll finally explore the data in time. The data is already loaded in the file widget. As we're not interested in latitude and longitude at the moment, we'll use Select columns to put them into MetaColumns, and thus exclude them from any calculations. To simplify the analysis, we'll select some countries we are interested in in the data table. I'll choose Italy, US, Iran, France, and the Chinese province, Hubei. As before, we'll use transpose to turn rows into columns and vice versa. We want to have countries as columns, while our dates will become a single time variable. Now our data is ready for time series analysis. Let's plot it with the line chart. Each curve on the plot is a country. We see that the virus started its path in China, then made its way to other countries. The slopes of the curves tell us how fast the numbers are growing, but they are very different in scale. We can log transform the y-axis for easier comparison. Plots show us that different countries came in contact with the virus at different times. It would be much easier to compare them if curves started their growth at the same time. To align the curves, we'll need some Python magic. Copy and paste a few lines of code into the Python script widget and run it. The script is available in the description below. With it, we have aligned the curves to the moment where the country recorded 100 cases. Let's take a look at how fast the virus is spreading. We'll use the difference widget. For each data point, the difference widget computes the change over a given time period. In our case, the difference in a day. Differencing order of one means we'll be looking at a derivative of first order, which means we'll identify critical points in our data, that is jumps and drops. Let's look at, for example, China and its transformed version. It's easy to spot the daily spikes. The most prominent spike here is due to the adjustment of counting. We can also compare countries and see how well they were able to contain the spread of the virus. The difference widget also has the option to output the percentage of change. This will allow us to observe the relative growth and compare the countries directly. Let's look at, for example, the Chinese Hubei versus France. It seems China is doing worse at the beginning and better later on, but it is impossible to tell at which point the trends start to shift. There are just too many jumps due to noise in testing and reporting. So for our final step, let's smooth things out. Smoothing is usually a part of pre-processing, so insert the moving transform widget between S-time series and difference. Then add a moving average transform, say a mean of every five values in the series. The trends now seem clearer. The countries have somewhat similar trends up to the 15th day, when China's growth falls decisively below the French. Curves benefited from smoothing and so did the difference plot. As a final step, we will add a difference widget after moving transform and plot the smooth difference between the two countries. Unlike the previous plot with the percentage of change, this will show us the absolute growth. The turning point seems to be on the 25th day. This was a longer video where we explored how to plot and transform time series. We learned how to align the curves, scale, difference and smooth them, with each transform giving us a different type of information. For now, this concludes our COVID-19 series. We hope that you feel empowered to grab the data and explore it yourself.