Informed debate on climate change is hindered by naïve interpretation of trend data. I have more confidence in the understanding of stock-market traders than in that of most climate change commentators
The more I look at the arguments on climate change, the more concerned I am at the crude misapplication of statistical methods that I find in them. This note is at least as reliable as views being expressed by many major contributors to the climate debate.
Lies, damn lies and crappy statistics
How much faith do you have in statistics? If you have formed a view on climate change, you should at least also have a view on what can and can not be concluded from a statistical examination of a database. And on the credibility of the analyses made by other commentators. My position is this. Statistics are generally considered to be the collection, analysis and interpretation of data applying various mathematically derived methods. That will do as a working definition. Data is plural of a datum. Data are the bits of stuff collected – the daily temperature readings from a weather station for example, or for a patient in hospital, or the price of a Corporate stock. As a matter of fact all of these data look rather similar when visualized as a time series.
We are pretty sure that the fluctuations of the patient’s temperature will ‘spike’ but in general will return to a standard basal level. Stock-prices as we know too well are less predictable. And as for charts of data about climate change, well these are in a different class altogether.
The First law of statistics
The first law of statistics for me is that statistical interpretation has to follow statistical understanding. Opinion is OK, informed judgement is different from opinions. Public debate tends respects the rights of individuals to express honestly expressed views. Phone-in programmes and the majority of blogs are not much more than that.
The second law of statistics
There is no safety in numbers if you don’t know what they mean. This brings me to one difference between climate debate numbers and those stock market and patient temperature charts. Climate debate numbers are increasingly highly complex abstractions which have been developed to examine a theoretical model.
The third law of statistics
Correlation is not the same as causality. My old statistics teacher used to quote the example of storks found perching on houses with the greater number of children born in them in one study. There was a correlation (perhaps because big warm houses attracted storks) but not one which can confirm a causal relationship between stork presence and human birth data.
The fourth law of statistics
Trend lines are treacherous.
I base this on the mathematical fact that there are a very large number of ways of drawing a line through thirty of so data-points. The zigzags of the data points are smoothed out in all sorts of cunning ways. The simplest smoothing is a straight line. It invites you to decide what all the deviations from that straight line mean.
A simple illustration
I have been playing around with a simple way of visualizing trend data. Let’s take the yearly data on temperature changes over a thirty year time period, the recent battle-ground in the climate debate
Many charts have appeared showing a trend as a straight line (so we have to beware of the fourth law). I find the following little thought-experiment revealing. Take a thirty point trend chart and select the highest and lowest items on the chart.
Mentally add each point in turn as the first datum of the chart. Almost always, one or both new visualizations change your perspective of the trend line and where the chart is going into the future.
Now repeat the experiment with each of the same two points in turn at the end of the trend chart. You will again find the visualizations offering one or maybe two fresh perspectives
Doesn’t that just confuse the issue?
Maybe it does. It weakens confidence in just what the trend-line might be telling you. But perhaps a bit more confusion and a bit less conviction is what is needed at the moment. Anyway, I do hope you will be able to contribute to more productive discussions on climate change in future.
I am no Fellow of the Royal Society of Statisticians, but I check my views from time to time with someone who is. The post captures my beliefs as a relative outsider to the Climate debate. I am suggesting the ‘laws of statistics’ in the sense of guiding principles on which I develop my case, rather than universal truths. Nor am I suggesting that visual inspection of trends is a substitute for careful application of statistical testing. But developing skills of visual inspection may enable more people to develop a sense of what a trend-curve might be signalling, and have a more informed discussion with those generating and interpreting such data.