Deltoid

It may be a parody, but I think you still made fewer errors and more sense than the average TCS column.

Posted by: dfinberg | January 28, 2005 12:32 AM

You mean, that "big dipper" graph actually appeared in a TCS article? Please tell me it didn't.

Posted by: Steve Reuland | January 28, 2005 2:15 AM

That didn't really help much. See, I used data all from the same set, and I get this <img src="http://vger.aa.washington.edu/triplett/webstuff/hockey2x_duh.jpg" width="100%" /> Looks like two hockey sticks to me. Are there really two hockey sticks? Why or why not? Was that first hockey stick due to man made causes? Maybe it's from the "little industrial revolution" of 500 c.e. or something. Now then, if it's fair to smooth out the old stuff to a greater degree than the new stuff becaues we have better measurements for the new stuff, that means that you can't compare non-smoothed features in the new data to smoothed features in the old data. It's not a fair comparison. The smoothing of the old stuff destroys the features that it might have had that we could have compared with the new stuff, no? If I'm wrong, why? Isn't the hockey stick supposedly showing warming unprecedented in the last 2000 years? I don't get it.

Posted by: ben | January 28, 2005 11:49 AM

hmmm, I guess pics aren't allowed. It showed up when I previewed the post.

Posted by: ben | January 28, 2005 11:56 AM

Ben, First, eyeballing and drawing in 'best fit' lines is a bad idea- it's really easy to substitute what you think you know for what the data actually tell you. And this picture is a great example of that, there's no way that the first 'hockey stick' is a good fit for that data... [Tim wrote something about this during the election that touches on this topic : http://tinyurl.com/5njnv ]
Second, you're asking a lot of good questions, but they are 'how do statistics and climatology work' questions. That is, they aren't going be answered in a sentence or two, you're going to have to do some real studying to figure them out.
For example, do you understand what Kieran means by confidence intervals? The data aren't being smoothed in the sense of rounding, so there isn't information being "destroyed"... we've just got more accurate measurements in the very recent past, so the uncertainties are smaller.
To answer another part of your question, a great deal of the modeling that goes into these graphs involves correlating data sources with each other. For example, if we were using tree rings as a proxy measurement for rainfall, we could examine recent rainfall records and recent tree cores to make sure that we are indeed measuring apples and apples (ie if our tree ring-rainfall inference/model is incorrect, we might expect our calculated results from modern tree rings to differ from the very reliable meaures of modern rainfall).
Go check out www.realclimate.org; they're dedicated to climate science, and so far the message board denizens have been very good about offering explainations.

Posted by: Carleton Wu | January 28, 2005 2:44 PM

Carleton, I know that eyeballing a best fit is dumb. I did it to make a point, which you acknowledged in your reply. I know the first hockey stick is no good, but for that set of data, neither is the second. The second hockey stick is supported by modern data, and that's all well and good, but you can't say that there wasn't a hockey stick in the past, because you don't know. So the hockey stick graph doesn't show anything historically unique, except maybe increased accuracy in temperature readings with time.
Please someone tell me you know what I'm trying to say. Don't refute points that I'm not making.

Posted by: ben | January 28, 2005 11:33 PM

...but it's very, very, very unlikely.

why?

Posted by: ben | January 30, 2005 2:45 AM

The reason that it's very, very, very unlikely has to do with statistics. Let's say that you took a group of 100 people, and selected 10 at random. You measure the heights of those 10 people, and figure that the average height of the group is 5'8". But how sure are you that this is the correct number? Better, how sure are you that the actual average of the entire group is between, say, 5'7" and 5'9"? Given certain assumptions (eg random selection, regular distribution (ie the class isn't made up of 95 pygmies and 5 NBA centers)), one can calculate the probability that the average does indeed fall within the range. Of course, there is a small chance that you've picked the 10 shortest or tallest people in the room- ergo, the associated probability.
So, it's possible in that scenario that the average height is actually 6'2", but it is very unlikely.

If you'll follow the link I provided earlier, you'll see several diagrams of the Mann and Jones 2003 (NH) data with the associated uncertainties. Bearing in mind that the actual line could lie outside of the uncertainty range (ergo, there could be a hockey stick back there someplace)- but it probably doesn't.
Also see http://www.realclimate.org/index.php?p=11 for their "myths about the hockey stick". If we only had one data set that supported the idea, the scientific conclusion would be much more tentative. But it's been supported by every peer-reviewed subsequent study, and a number of other analyses- so scientists are much more confident about the conclusion.

Posted by: Carleton Wu | January 30, 2005 10:35 PM

Do you actually understand the effects of the mathematical contortions Mann went through to produce the "stick" ?

WOW, do you actually understand the term "proxy"? No? I thought not.

This is a dead parrot. It has gone to meet its maker.

D

Posted by: Dano | February 4, 2005 12:58 PM