The Hottest Day of the Week
Are Sundays actually hotter than Fridays? 🌡
I was looking at a chart presenting temperature data for Colombo for the last couple of months, and I noticed something. For some reason, Fridays seemed to be cooler than other days, and Sundays hotter.
I wondered if my eyes were playing tricks, and so I thought of plotting the mean temperature for each Day of the Week for Colombo. This is what I got:
Sure enough, Fridays seemed cooler than Sundays.
Then I wondered if other parts of Sri Lanka displayed similar patterns. The results were even more intriguing:
In Galle, the heat seemed to spike on Saturday, and then crash on Sunday.
In Jaffna, the middle of the week was cool, and then the temperature rose, again peaking on Sundays.
In Kandy (weather station is in its suburb Katugastota), the pattern was similar to Jaffna.
What was going on? What caused all of this?
Can you guess?
Let me give you a clue: All of the patterns above had the same cause.
Answer: Randomness. The patterns above were all caused by pure chance.
How do I know this? Simple. I plotted the lines above, with the range in which mean temperatures tended to vary (also known as “Variance”).
The pink region colors a range of two standard deviations.
If we are to conclude that mean temperatures on two days of the week (e.g. Sunday and Friday) are actually different, then there must be zero overlap in the pink regions.
As you can see, no two days for any of the places we looked at have zero overlap. In fact, the overlap is almost complete, meaning that the temperature is almost identical across the days of week.
I can sense that you’re a little disappointed, even betrayed. You came in search of an interesting nugget of data, and now you have to leave, empty-handed.
Well, you won’t leave empty-handed. Because the real reason I penned this article was to make a point, which has nothing to do with the weather.
Don’t take charts in particular, or statistics in general, at face value. It is very easy to mislead and be misled by them. Always dive deeper.
Whenever you need to make an important decision (e.g. buy a stock on the stock market, pick one candidate over another, pick a location to build a house) and you are guided (or think you are guided) by data, always check if the data is actually meaningful. In particular, whether any “trends” however interesting and attractive them might look, are caused by pure chance.
