Shades of upside down Tiljander. McIntyre is delving further into the Marcott proxy issue and it looks almost certain now there’s a statistical processing error (selection bias).
Steve McIntyre writes:
In the graphic below, I’ve plotted Marcott’s NHX reconstruction against an emulation (weighting by latitude and gridcell as described in script) using proxies with published dates rather than Marcott dates. (I am using this version because it illustrates the uptick using Marcott methodology. Marcott re-dating is an important issue that I will return to.) The uptick in the emulation occurs in 2000 rather than 1940; the slight offset makes it discernible for sharp eyes below.
Marcottian
uptricksupticks arise because of proxy inconsistency: one (or two) proxies have different signs or quantities than the larger population, but continue one step longer. This is also the reason why the effect is mitigated in the infilled variation. In principle, downticks can also occur – a matter that will be covered in my next post which will probably be on the relationship between Marcottian re-dating and upticks.
Read his full post here: How Marcottian Upticks Arise
Maybe we need an Uptick Rule for paleoclimatology
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Here you go Willis
Just plot these two graphs and then tell me you can’t see a relationship.
First graph
Cycle average Maximum
13.8
12.7
15.0
15.6
13.7
Cycle toal sunspots divided by 1000
24.4
21.1
30.8
29.0
19.8
Second graph
Cycle average Mimimum
5.6
5.4
6.3
6.8
6.2
Cycle total sunspots divided by 10000
2.4
2.1
3.1
2.9
2.0
Best Regards Kelvin
I wonder how much of the apparent problems that stem from statistical issues arise out of how easy it is to undertake complex and subtle analyses nowadays without knowing much about what is actually being done.
Kelvin Vaughan says:
March 18, 2013 at 3:33 am
OK. I plotted the graphs, and I can’t see a relationship.
Or, since all things are related in some way, let me be more clear.
There is no statistically significant relationship in either graph.
Kelvin, you seem like a nice guy … but you desperately need a class in statistics. More men than you have been fooled by apparent relationships that are not significant in the slightest. That’s why we have statistics …
w.
Willis Eschenbach says:
March 18, 2013 at 8:35 am
Thanks Willis.
I went and searched my copy of excel and found a correlation formula which I have been playing with comparing total sunspots in a cycle and total yearly temperature in a cycle.
I have now checked data from three stations and found correlations rangeing from -0.99 to -0.77 and from 0.92 to 0.64. I found that if you correlate years prior to the cycle ending in 1964 you get a positive correlation and for the 1964 cycle onwards you get a positive corellation, at all stations. ?????
The best correlation was for Cambridge UK. -0.99 and -0.97 for max and min since its inception in 1959.
regards
Kelvin
Kelvin Vaughan says:
March 19, 2013 at 10:07 am
Kelvin, that’s good stuff, and there’s a couple of problems with that.
First, you have to figure the odds that you got the result by random chance. This is commonly called the “p-value”.
Next, you need to adjust your figures for autocorrelation.
Next, if you look at say a dozen stations and you find one with a p-value of 0.05, is that significant? If that were the first and only station you’d looked at, sure.
But if you look at a dozen of them, you have about a fifty-fifty chance of finding one with a p-value of 0.05.
In any case, I’m overjoyed that you are out there trying to find correlations … you just need to be careful, there’s lots of pseudo-correlations out there that disappear once more data is available.
w.
Thanks Willis
I will look up p value next. I am getting such a good correleations that I think I must be doing something wrong.
Thanks for the tips. This is keeping me occupied in my retirement.