6TH OR 7TH HOTTEST APRIL IN 125 YEARS!

That's not what he's saying. He's trying to show an exponential fit because 'exponential' sounds much worse than linear. It isn't enough that temperatures are rising rapidly. They must be rising exponentially!! There is no climatological reason for temperatures to be increasing exponentially except in Roger's own mind. It's the distinction between modeling and fitting. Fitting is a very crude process, as you can see from Roger's efforts. There are benefits to it, but at the end of the day you don't learn anything about cause and effect. Modeling requires a much more rigorous approach, including an understanding of the reasons that the data takes the shape it does. Without that understanding the results are from such an exercise are limited. Roger has been asked repeatedly to explain why one would expect an exponential increase in temperature and to date has failed to come up with a compelling reason that it should be that way.

A least squares fit is the standard method used to estimate and minimize the error between a known mathematical function and a set of data.

The term 'squares' is used because the technique estimates error based on the square of the difference between each data point and the computed data point for the same domain as the known data point.

The square is used in part because it is easy to compute, provides a positive result that can be summed even if the difference between the data and computed points are negative, and because there are strong connections between this form of error measure and the concept of 'standard deviation'.

Basically, for every point in the data set (x coord), the proposed function is evaluated (at x) and a new y coordinate is generated. This coordinate is then subtracted from the datum's y coordinate and the result squared. The squares are then summed over all data points.

The term 'least' is used because in the case of curve fitting, the mathematical procedure used selects the function that 'best fits' the data through the minimization of the sum of the squared errors computed as indicated above.

Yes, the models show acceleration also. I have two reasons to believe this. The first Is that I've watched the difference in residules between the linear fit and the exponential fits. The second is that I have noticed that in this section of my report

The global mean temperatures of the recent months are almost always above what would be expected by linear projection.

My days of teaching statistics are over. (Or more simply, 'My days are over.' I don't have the time to teach you, (or Ian and David). See if you can enroll in a statistics class somewhere. There may also be some on-line instruction on the net.

I agree with this. If all you want to do is drive a line through data to estimate the past influence of an 'independent' variable on a 'dependent variable', curve fitting is a good method. What curve fitting is not good at is extrapolation. Roger doesn't realize that the least squares line has hyperbolic confidence intervals around the data. This means that if you extrapolate to any significant degree, your confidence in the extrapolation rapidly approaches zero!

One other problem with Roger's efforts: correlation is NOT causation. A curve fit will tell you that two variables are correlated, but it will not tell you that one variable causes the other. In other words, in many countries, a plot of soft drink sales vs. new malaria cases will indicate that soft drinks cause malaria, since booming soft drink sales come slightly before much higher malaria cases. The 'hidden' variable in this sad story is ambient temperature, which is the true 'driver' for both of the other variables.

He can't come up with this reason because 'that ain't the way it is'.

Thanks, everybody. I think I understand Roger's post better now

Which is why I made no causative claims.

(cut)

I don't think you need to worry about it. With the way that the economy, the war, the budget deficit, etc., have been going, it is very likely that you need to worry about what the next President thinks.

There are a few fossil fool flunkys here who would disagree, but it is unlikely that they are reading your comments, and even less likely that they will reply.

In two or three decades, my statemests about exponential global mean temperature growth will be tested. Mark my words well.

Well, I've only got a 125 year data set. To map the mindset inside the Bush White House, I would have to have data going back to the Bronze Age. I am sad to say that there is no hope for this, my computer system does not have a clay tablet reader.

That he or she thinks would be a good thing.

josh halpern

Roger, you're familiar with Leon Festinger's 'theory of cognitive dissonance,' aren't you?

I haven't actually read Festinger's original theory and the research supporting it, but a prof summarized it for me in an environmental psychology class back in graduate school.

Basically, Festinger theorized that it's mentally painful for people to hold two sharply clashing ideas in their minds at the same time, at least if they care emotionally about the implications of either idea. And so Festinger theorized that most people will find SOME way

On 12 May 2004 11:16:22 -0700,

Nice. Anyone who would disagree is a fool, a flunky, and not thinking. I have got to remember that debating tool.

Hint: not everyone agrees with the notion that all global warming is anthropogenically caused.

Of course, trying to debate that with someone like Roger is a waste of time and it annoys Roger.