Ignoring Statistical Significance to Promote Climate Alarmism
Each day I read new pieces in the mainstream media on climate change, and I continue to be amazed by the poor quality of science journalism and how this is misleading the public on critical policy issues.
Seth Borenstein of the the Associated Press has published a truly staggering piece on temperature trends across the United States over the past three decades that is getting picked up by various media outlets around the world. I simply cannot believe that the scientific community is allowing this article to stand without major correction or a full retraction. The end days for science are upon us.
Borenstein claims the following methods of analysis for his article:
To determine what parts of the country have warmed the most, The Associated Press analyzed National Climatic Data Center temperature trends in the lower 48 states, 192 cities and 344 smaller regions within the states. Climate scientists suggested 1984 as a starting date because 30 years is a commonly used time period and 1984, which had an average temperature, is not a cherry-picked year to skew a trend either way. The trend was calculated by the NCDC using the least squares regression method, which is a standard statistical tool.
Then he claims the following findings:
All but one of the lower 48 states have warmed since 1984. North Dakota is the lone outlier, and cooled slightly. Ten states – Maine, Vermont, New Jersey, Massachusetts, New Hampshire, Rhode Island, Delaware, New Mexico, Connecticut and New York – have gotten at least 2 degrees warmer in the past 30 years.
Since 1984, 92 percent of the more than 500 cities and smaller regions within states have warmed and nearly two-thirds of them have warmed by at least a degree. The regions that have warmed the most have been New York's St. Lawrence Valley, northeastern Vermont, northern Maine, the northeastern plains of New Mexico and western Vermont, all of which have warmed by more than 2.5 degrees.
Here is what happened. The AP used "the least squares regression method" to calculate the annual temperature trend for all these regions, but then proceeded to ignore entirely whether the regression method indicated if the trend was statistically significant (the typical criteria would be a p-value<0.05).
This is first-year statistics level stuff. Quite simply, if your statistical test ("least squares regression method") tells you the trend isn't significant, you cannot claim there is a trend, since the null hypothesis (i.e., no trend) cannot be rejected with any reasonable degree of confidence.
This isn't the first time I've seen statistical troubles from this journalism source. In August 2012, Borenstein attempted to analogize the odds of consecutive warm months to coin-flipping statistics, which – as I noted – was not a valid analogy, as it violated the requirement of independent events.
To correct Borenstein's reporting on this latest topic, I analyzed the 1984-2013 period of annual temperatures for "the lower 48 states" using "the least squares regression method," as the AP did. Here is what I found.
Borenstein claims that "all but one of the lower 48 states have warmed since 1984." Wrong. Only the following 18 states (i.e., 37 percent of the lower 48 states, not 98 percent of them, as the AP claims) exhibit statistically significant positive trends in annual temperature since 1984: Arizona, Colorado, Connecticut, Delaware, Louisiana, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New Mexico, New York, Oklahoma, Pennsylvania, Rhode Island, Texas, Vermont, and Virginia.
To illustrate how scientifically ridiculous the AP's claims are, the AP article claims there is a cooling trend since 1984 in North Dakota and a warming trend in Washington over this same timeframe. Here is a plot of annual temperatures in these two states over the past 30 years.
You're kidding me. Claims that Washington is warming and North Dakota is cooling? The linear regression p-value is 0.94 for North Dakota and 0.66 for Washington. Keep in mind for statistical significance the p-values must be below 0.05 (or even 0.01 if you wish to be truly rigorous). Both these states show almost perfect non-correlations. In other words, there is no hope of a temperature trend (towards either warming or cooling) in either state, contrary to what the AP is claiming.
Of the nine climate regions constituting the contiguous USA – the Ohio Valley, Upper Midwest, Northeast, Northwest, South, Southeast, Southwest, West, and the Northern Rockies and Plains – only the Northeast, South, and Southwest have significant increasing annual temperature trends over the past 30 years.
And as I have previously shown, only the Northeast climate region has a significant warming trend since 1991. No other climate regions do, nor does the contiguous U.S. as a whole. Thus, the results from the 30-year period beginning in 1984 are not statistically robust, as the statistical significance for trends in two of the nine climate regions disappear after only the first several years.
Nor does it matter whether or not 1984 "had an average temperature" when we assess whether the data has been cherry-picked or not. Furthermore, the 30-year long "commonly used time period" in climate science certainly reeks of arbitrariness. Why 30 years? Why not 29? or 31? or 34? or ...you get the picture. Just because a group of scientists decide on a common timeframe to use for their analyses does not make it any less arbitrary and subject to valid accusations of cherry-picking.
As for the AP's claims that "since 1984, 92 percent of the more than 500 cities and smaller regions within states have warmed," my analysis of trends since 1984 among the 338 individual climate subdivisions for the United States shows that only 99 have statistically significant warming trends. That is only 29 percent. Among the 186 cities, 80 did not have complete climate records since 1984 and were omitted from future analysis (who knows how the AP dealt with these incomplete records?). For the remaining 86 with complete three-decade long records, only 32 (or 37 percent) had significant warming trends.
Overall, only 31 percent of the "cities and smaller regions within states" have significant warming trends since 1984. This is a long way from the AP's claims of 92 percent, isn't it?
So this is how climate science works today? Completely ignoring the basic requirement for statistical significance during trend analysis in order to promote an alarmist agenda? What complete and utter scientific rubbish and exceedingly poor science journalism. A War on Science indeed, except it is from the climate alarmists.