Little Fish, on 24 March 2013 - 10:54 PM, said:

*"there has been a consistent rise in temperature over the last 15years" when a statistically meaningful analysis is carried out."*

can you show a statistical meaningful analysis showing the temperature trend consistently rising over the last 15 years.

At 95% confidence, the temperature rose at an average rate of 0.012853 degrees per year. That's a trend line. The straight-line model accounted for 36.6% of the variation and produced an F-value of 8.07 with 1 and 14 degrees of freedom. Probability of the trend line actually being flat: 1.31%. That's significant.

But, any analysis of current temperature trends that is less than 30 years long is cherry-picking. In climatology you don't get to pick the interval; 30 years is the specified minimum length of a trend line. The 30-year line (1983 to 2012) produced these results:

At 95% confidence, the temperature rose at an average rate of 0.016754 degrees per year. That, too, is a trend line. The model accounted for 72.6% of the variation and produced an F-value of 74.04 with 1 and 28 degrees of freedom. Probability of the trend line being flat is less than 1 in 10,000 (The program won't calculate a number that small.).

There is too much variability during the last 15 years (1998 to 2012) to calculate a trend line. This means that Little Fish's argument boils down to: "At 95% confidence, there is insufficient evidence to conclude that temperature is changing, so that proves it isn't." Does anybody see the logical problem here?

Because the 1996 to 2011 model accounts for more variation, it is a better description of reality. But because the 1983 model does a job several times better, it is an even better description of reality.

The argument about whether the temperature trend has changed in the last fifteen years is couched in a partial truth, no matter which side of it you're on.

Doug

**Edited by Doug1o29, 25 March 2013 - 01:56 PM.**