[asa] On Error Bars and Ordering

From: Rich Blinne <rich.blinne@gmail.com>
Date: Mon Dec 21 2009 - 00:28:57 EST

One of the fundamental differences between Glenn and myself deals with the issue of error bars and ordering. If two different data sets have overlapping error bars I'm happy even if the ordering of the data points are different. For Glenn, this is a sign of hopelessness and incompetence. I believe a simple example is in order, the political poll. There is only one winner and one loser at any one time. Let's compare two polls:

Poll 1:

Candidate A: 49%
Candidate B: 51% MoE 4%

Poll 2:
Candidate A: 52%
Candidate B: 48% MoE 4%

But only one candidate can be winning. The pollsters must be wrong. Pollsters are probably the only group of scientists -- yes I call them scientists -- more maligned than climatologists. Others might say that by averaging the two polls that candidate A is winning. The pollsters say no, no, no it's a statistical tie. What the pollsters want to see is a lead larger than the margin of error and at that point the ordering has significance.

Now go to the AGW sweepstakes. Who's the hottest year of all? Earlier today I posted today an ordering by GISS that 2009 was the fourth hottest year. IIRC NCDC declared it number 5. What gives? Let's take a look at the graph with the error bars.

http://data.giss.nasa.gov/gistemp/graphs/Fig.A2.lrg.gif

The green bar is the margin of error. Any of 1998, 2002, 2005, 2007, 2009 can be in any order and any of them can be the "winner". When the different climatological organizations construct their top ten lists the ordering can and have been subtly different because of this. And as the polling example shows averaging the different results does not improve our knowledge of the ordering and many times has a confounding effect. Just because there has to be a real historical ordering we may not know it. The perfect becomes the enemy of the good. Furthermore with a 0.1 degrees C error bar and an average of 0.15 degrees trend line any trend less than a decade is statistically meaningless. As we have experienced warming since 1980 the IPCC has concluded that AGW was real but they did that with greater and greater certainty as the trend has escaped the error bars. Most people only care about the data points but where most of the scientific action is is trying to reduce the error bar!
 s. One of the great frustrations of climate change science is that while we have a well-understood and a solid range of climate sensitivity the errors bars have been frustratingly static over the last couple of decades. The wide spread makes it difficult to plan what is a safe level of CO2 because of it. Policy makers want a single number but the climatologists cannot give them one. People complain about the computer models being uncertain but it's the uncertainty of the climate sensitivity that makes the trade off between mitigation and adaptation hard.

Rich Blinne
Member ASA

To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Mon Dec 21 00:29:19 2009

This archive was generated by hypermail 2.1.8 : Mon Dec 21 2009 - 00:29:21 EST