Statistical Analyses of Surface Temperatures in AR5（统计分析AR5的表面温度）.pdfVIP

Statistical Analyses of Surface Temperatures in the IPCC Fifth Assessment Report Douglas J. Keenan 17 December 2013 First version 28 December 2016 This version Acknowledgements: for comments on drafts, I am grateful to Sir David Cox, David Henderson, Demetris Koutsoyiannis, Janice Moore, and William Nordhaus. This critique is available at /AR5stat.pdf 1. Introduction In September 2013, the IPCC issued the first volume of its Fifth Assessment Report (AR5). The volume includes extensive discussions of observations of the average temperature on Earth’s surface (i.e. where people live). The discussions include some statistical analyses of those observations. This critique considers the merits of such statistical analyses. No background in statistics is required. 2. How to do a statistical analysis To understand statistical analysis, consider an example . Suppose that we toss two coins, and we ask what the probability is that both coins come up heads . To determine the probability, we will make two reasonable assumptions: (i) the probability that a coin comes up heads is ½ and (ii) one toss has no effect on the other toss. Then, the probability of two heads is calculated to be ½ × ½ = ¼. In general, any statistical analysis will consist of two phases. The first phase is to make some assumptions about the process that generates the data (in our example, we made two such assumptions). The second phase is to do some mathematical calculations (in our example, we did a simple multiplication). The assumptions are obviously vital for the analysis. If, for instance, we had assumed that the probability a coin comes up heads is 1/3, then our final answer would have been different: 1/9. The assumptions, collectively, are called the “statistical model” . Every statistical analysis thus vitally depends on what statistical model—i.e. set of assumptions—is chosen. Although the mathemat