Thursday, 29 September 2011


So it has been warm lately hasn’t it? The only reason I know this is that twitter and facebook have been full of people mentioning it. As usual this year, whenever there is good weather I am stuck in an office where even the concept of Summer is hard to grasp. Sunshine outside might as well be sunshine on Mars for all the satisfaction I get from it while I toil.

My usual reaction, on hearing how much people are enjoying something that is denied to me, is one of dismissive envy. In this instance this manifested itself in a belief that actually this warm spell is not at all remarkable anyway - that people are disorientated by the piss-poor Summer just passed and are just blindly seizing on the opportunity to moan about how hot it is for the first time this year. I hypothesised that people were relying on flawed memories for their perception of what the weather was meant to be like at this time of year. You know, like in all that research that shows how fallible eye-witness statements are? That’s exactly what’s going on, I thought.

Having nothing better to do, I thought it would be fun to get all the data for maximum temperatures in September over the last decade and compare them in a graph. Then I could show people that while it is indeed warm it is not remarkably so. I’d show them that actually it’s often like this and they would concede that there was no point banging on about it as they had been doing all week.

So I went to the excellent and copied all the data for the Septembers over the last 10 years at Gatwick airport (nearest data to where I live in Hove), right up until yesterday. Using my rudimentary I.T. skills, I put all this stuff on a graph and here it is:

[Hmm, it's a bit small but the X axis is the day of the month of September (1-30), the Y axis is the maximum daily temperature in degrees centigrade (10-30). On the right is the key showing the colours allocated to each year. The RED line is the one for THIS September - the red line ends yesterday]

As you can see, the bright red line, which shows maximum daily temperature for each day in September 2011 (that’s the September we’re in now), is actually significantly higher right now than it has been on the same day over the last ten years. My hypothesis is therefore wrong. The people were right, after all. They were right to say that it is warmer than usual and I was wrong to dismiss this as some kind of mass delusion - wrong to assume that they had forgotten similar temperatures on this day in previous Septembers.

When this result emerged on my graph, I was disappointed and initially I thought “oh well, I won’t write that blog entry after all, my hypothesis was wrong” but then, of course, I would have failed humanity! I must share the result of my investigation whether it proves me right or not!

In this instance, my efforts have proven my reaction to be wrong but that is not a bad thing. I have now learned to put a bit less faith in my grumpy knee-jerk reactions. Next time I respond to such a situation with grumpy dismissal, I am more likely to remember the time I proved this reaction to be unreliable. This will result in me getting over grumpiness more quickly, armed with a new tool to dispatch it with. Thank you skepticism!

[I suspect that this conclusion would be less clear if I went back 50 or 100 years, as weather really does fluctuate. However, my knee-jerk reaction against the “ooh it’s remarkably warm” comment was based on my lack of faith in people’s memories of the September weather over the last few years. Also, when I say that this one is “Significantly” warmer than Septembers over the previous decade I mean this in a purely intuitive sense - I don’t know enough maths to say that it is actually significant in a truly mathematical sense. Maybe it is.

You can see that 2003 and 2006 also had unusually high maximum temperatures, but these were earlier in September when the average temperature is slightly higher than it is at the end of September. These periods would also have been “remarkable” by my definition, but less remarkable than this warm spell we’re in right now.]


  1. It's great to see someone publishing info that goes against their initial hypotheses. We don't see that enough in the scientific community. Acocrding to a Skeptics with a K podcast I listened to recently, apparently many journals don't publish repeated studies, nor negative findings. Have you thought about doing some stats training? It can be a little dry at times, but if you've not read stuff by Ben Goldacre (I'm sure you have) he makes it quite accessible. I did some at uni and once you understand the basics, it's great when evaluating studies. In fact, I was actually thinking about blogging for a lay audience about study evaluation. I may think some more about that now. Thanks for reminding me!

  2. Thanks Tannice, that is a bloody good idea! I learnt it properly ages ago but the knowledge has leaked away through lack of use. I ought to sort that out really.
    I did indeed read and love Bad Science!

  3. Funnily enough, Ben writes about this kind of thing in his column today (tomorrow)!