Mann, Bradley and Hughes have published some corrections to the supplementary information for the famous hockey stick graph showing the temperature record of the last 1000 years. They say that the errors do not affect their published results. This could explain why McKitrick and McIntyre could not reproduce their results, but McKitrick is continuing to insist that Mann’s graph is wrong.

McKitrick has also published some errata. Unlike Mann’s error McKitrick’s error affects his results:

Figure 3 in the Cooler Heads Briefing on TBS contains an error. Tim Lambert of Australia has pointed out that missing data were handled differently between Figures 2 and 3, and when this is fixed the example no longer illustrates the intended point. The point (that the trend can change if the averaging rule is changed) is shown in this Revised Spreadsheet. Our thanks to Tim Lambert for pointing out the error.

(The post where I pointed out the error is here.)

I looked at his revised spreadsheet. This time he has dealt with missing values consistently and it does indeed show a warming trend when the usual arithmetic mean is used and a cooling trend when their unusual root-mean square is used. So how did he manage this? After all, as I showed in my earlier post, the root-mean square in Kelvins gives almost he same answer as the regular average. Well, **McKitrick invented his own temperature scale**. McKitrick modestly did not give it a name, but I am dubbing it the McKitrick scale in honour of its creator. To help you gain familiarity with this new scale, the form below lets you convert between degrees McKitrick and the old-fashioned degrees Celsius and degrees Fahrenheit. Just type a number into any of the boxes and press “Enter”.

Anyway, in his revised spreadsheet McKitrick takes the root-mean-square average of temperatures measured in degrees McKitrick. This way of averaging temperatures gives some rather odd results. For example, the RMS average of -10°M and -10°M is not -10°M as you might expect, but +10°M. Needless to say no-one actually uses RMS averages of temperatures in the McKitrick or any other scale, and no-one in their right mind would use them.

So revising their original example to use degrees McKitrick means the trend is different for different averaging methods? Well, no. If you take their original example and use the root-mean-square-in-degrees-McKitrick average, you still get the same trend. In the revised spreadsheet McKitrick has also changed the set of weather stations used. Even then it makes little difference to the size of trend—it changes an insignificant warming trend to an insignificant cooling trend.

To summarize: even if you use a weird root-mean-square-in-degrees-McKitrick average it makes little difference to the size of any warming or cooling trend you might see.