An amusing exchange of views between Bob Ward and Phil Jones on statistics:
I was wondering whether you have seen the article by David Whitehouse in the latest edition of 'New Statesman'? http://www.newstatesman.com/200712190004
It would be great if somebody could respond to the article. I would be happy to do so if somebody can supply me with the ammunition. Any thoughts?
Jones responds swiftly:
From: Phil Jones
Sent: 20 December 2007 13:58
To: Bob Ward
Subject: Re: More nonsense on climate change
Quickly re-reading this it sounds as though I'm getting at you. I'm not - just at the idiots who continue to spout this nonsense. It isn't an issue with climatologists. All understand. If I tried to publish this I would be told by my peers it was obvious and banal. I will try and hide it in a paper at some point. I could put it on the CRU web site. I'll see how I feel after the Christmas Pud.
I would have thought that this writer would have know better! I keep on seeing people saying this same stupid thing. I'm not adept enough (totally inept) with excel to do this now as no-one who knows how to is here.
What you have to do is to take the numbers in column C (the years) and then those in D (the anomalies for each year), plot them and then work out the linear trend. The slope is upwards. I had someone do this in early 2006, and the trend was upwards then. It will be now. Trend won't be statistically significant, but the trend is up.
This is a linear trend - least squares. This is how statisticians work out trends. They don't just look at the series. The simpler way is to just look at the data. The warmest year is 1998 with 0.526. All years since 2001 have been above 0.4. The only year before 2001 that was above this level was 1998. So 2cnd to 8th warmest years are 2001-2007
The reason 1998 was the warmest year was that it resulted from the largest El Nino event of the 20th century in 1997/8. We've not had anything resembling a major El Nino event since - they have all been minor.
Using regression, it is possible to take the El Nino event into account (with a regression based on the Southern Oscillation Index). This accounts for about 0.15 deg C of 1998's warmth. Without that 1998 would have been at about 0.38.
There is a lot of variability from year-to-year in global temperatures - even more in ones like CET. No-one should expect each year to be warmer than the previous. The 2000s will be warmer than the 1990s though. This is another way of pointing out what's wrong with their poor argument. The last comment about CET is wrong. 2007 will be among the top 10 warmest CET years - it will likely be 2cnd or 3rd.
Ward is impressed:
Thanks for responding so comprehensively. I have plotted the data before, and as you observe, the trend is up but the result isn't statistically significant, which I think makes it open to attack. I think the problem is that NOAA made the following statement in its report on the 2006 data:
"However, uncertainties in the global calculations due largely to gaps in data coverage make 2006 statistically indistinguishable from 2005 and several other recent warm years as shown by the error bars on the global time series."
I'm not sure how to argue against this point - it appears to imply that there is no statistically significant trend in the global temperature record over the past few years.