I suspect that as a population the world has never looked at charts as much as the last few months during the Covid 19 pandemic. Whether it is the stock market, your shares, your superannuation, your wage levels, health situation, petrol prices etc etc. Whenever we want to watch out for something that concerns us we would like to know a few things like:
- how are our current figures going
- how have they changed
- how are we going compared to someone else
- are things improving or getting worse over the last few days, weeks or months
- how does one thing compare to another
I worked for Government and later for private enterprise as a Senior Business Information Co-ordinator (long name I know) and a fair bit of that time was focused on business intelligence reporting. It was always hoped that it provided an “intelligent” view of what has happened and hopefully enabling modelling of what can happen in the future. Charts are something that every business intelligence person seems to love. They can be used to guide or misguide so must be understood, so therefor I will do my best to explain some things about them.
During the Covid 19 pandemic we have been very interested in “flattening the curve”. What on earth are they talking about? Well, if the curve is flat it is good and if it’s vertical, it is bad – ‘not helping’? The curve is a line or shape on the chart. A good test of a chart/curve is how easily it can be explained. If we are talking about something that is bad (like Covid 19 infection rates) a flat curve means infection rates are stretched over time or there are simply less occurrences, so flat is good because medical resources can cope better. Be aware though, that if you had a flat curve and it was showing that you have less superannuation than you have want, that is not so good. So flat and low curve results are good when talking about bad things.
Time to face what I might call chart shock
There are two “Confirmed cases” charts shown here from a popular web site using open source data. So the top chart is obviously not as flat and has gone up steeply and the bottom one is beautifully flat in comparison. Great! We might say “I hope the bottom one is the correct one and shows our situation”.
Now look at the dates and times on each chart and note that the date, time and figures are identical. Yes, they are showing exactly the same data, just viewed in a different sized browser window which changes the shape of the chart! So which one is correct? Well, both are. The top view is on a phone and the bottom on a PC. I think the top one is a more helpful shape because you can see changes easier and that’s what is important, but the data is exactly the same in the bottom one. I guess the lesson is to be cautious with charts as they are very helpful but can lead to the wrong conclusions. Also notice that in both charts the bottom line (X Axis) is not showing date but is showing how long since each country first had 100+ cases. This actually occurred at different dates for each country. So, the curve for each country has been brought together to show all countries starting at the same place on the left of the chart so they can more easily be compared with each other. The other adjustment is that the vertical figures (Y Axis) is not linear but more like logarithmic. So as the height in the chart increases, each horizontal line is at least double the amount of the one below it.
What you see may not be what you think you are seeing
Now let’s see another fairly surprising view of the same data. See the two following charts. I have included just three countries to make it less crowded but it has the same data in both charts. The data is displayed by date along the bottom but the big change is that the first chart shows heights on the chart representing the actual supplied data (linear height not logarithmic). The second chart below shows a logarithmic scale. The same as those popularly used on the web and the same height scale as the charts above.
Now, in the first chart the US looks to have reported so many more cases. Australia looks to have a very low number of reported cases. This is because the US did report a total of 842,629 cases by 23/04/2020 and Australia did report 6,654 cases.
It is just that the charts we often see have a logarithmic vertical scale (as in the second chart below) to allow the shapes of the curves to be compared. The height in those charts are quite misleading unless you know about the scale used.
So if we want to see Australia or other countries with relatively low figures on a chart, the tendency has been to modify how the chart looks so that you can compare the curves more clearly. It is a recognized technique, but can easily be misread.
What about incorrect data
Incorrect data can potentially cause some of the biggest problems simply because the data does not represent what happened. The chart below shows a large correction in some of the data. New data was reported at the end. Now the problem is where to put the data that was missed earlier. You see it placed at the end here which looks like the Covid 19 deaths are much worse on 22/04/2020 than they have ever been in China. We would hope this is not true. It is more likely that the new data should actually be distributed into the historical data. But that would change history as it was reported. In my experience, if there is found to be missing data in the past it is normally disturbed back in to historical data where it should be so that the chart becomes usable again and a note placed to explain what happened.
So what does all this mean
Data in itself is not usually easily read compared to a chart. Simply put, when you look at charts, it is very important to understand them whenever the subject matter is very important. They are powerful at illustrating data but they need to be read carefully. They may not even be showing some important data that could change your view of the information. By adding a new piece of information they can be suggesting that the new information relates to the existing information in the chart where it may not. A slightly funny example of this may be an old saying where someone says I don’t believe that an apple a day keeps the doctor away because my great Grandfather ate an apple a day and he is dead now. The misleading statement is still correct but he lived into his 90s. Who knows whether the apples helped.
Check the scales being used. For long term investments such as superannuation, be sure to look at all the long term data. To keep up with data relating to decisions you must make regularly (like when to fill the car with fuel) check the data regularly. Look for consistent patterns that relate to a relevant cycle. Don’t assume repeating cycles will always repeat.
Always get advice from a professional in the subject matter.