I’m interested in why you would assert this. I don’t doubt that you believe it but you really don’t know it to be true. This is your own invention, and in fact does not accord with what the authors themselves said they did, which was to review every paper that popped up in a search of “global warming” or “global climate change.” Sampling was absolutely not what they did.
No, you are right. This claim cannot be supported by the survey. The claim should properly be about the scientists, not the papers. Such a claim is supported by the survey, although I would allow a little more uncertainty in the conclusion, owing to the difficulty of classifying papers. I’m sure there were borderline cases whose disposition could easily be biased by the inclinations of the person doing the classification.
What I meant was they took all the papers that popped up in the search you just mentioned, and then classified them. The papers that took no position on AGW offered no information on what the authors believed, so they looked at the papers that did offer an opinion. This is the “sampling” I referred to. And in this very overt “sample”, they concluded that 97% of them supported AGW. That’s all. There was no secret sampling going on behind the scenes (that I know of). And there was no need for random sampling, because the population of papers was small enough that all relevant papers could be included.
Basically they excluded all but 3,800 papers from their sample of papers published over a ten year period and claim that proves 97% of scientists believe in man made global warming.
Coleman himself is actually very bogus. This can be found on Snopes and here is a line by line rebuttal from
There is nothing bogus about it. The papers that were excluded were papers that did not take a position on the question of man made global warming. There was no other criterion for exclusion.
So we can make claim that 97% of scientists believe in man made global warming based on the review of 3,800 papers published over a 10 year period?
My first impression was that it proved absolutely nothing except that oftentimes you find what you’re trying to find.
Yes. Statistical sampling is a well-developed theory that has been used for at least 200 years. What is so surprising about deriving a statistic for a population based on a sample? It is done all the time with exit polling to predict the outcome of elections with very good accuracy. Even if you only sample 1% of the population, provided the sampling methodology was not correlated with the characteristic you are trying to measure, the results of that 1% can often be applied with high confidence over the whole population. In the case in point, the sample included a third of the population. That is huge compared to other statistical samples. If 97% of the authors of sampled papers agreed with global warming, it would be quite surprising if only 50% or less of the authors whose papers were not sampled disagreed with global warming.
And what was the margin of error on this “sample” Was it a random sample? Why did they exclude those who said there wasn’t enough evidence to make a decision one way or the other?
Read the recent posts between me and Ender in this thread. They will explain how there was no need for a random sample because all relevant papers were considered, and it will also explain that papers that did not answer the question of global warming were not about answering that question. It would be a mistake to assume that every author of a paper that is indexed as a global warming topic was necessarily even about the question of global warming.
Again the selection process was only based on whether the paper was about this question. If it turns out that most people who answer this question answer it in the affirmative, that should tell you that maybe the correct answer is “yes”.
Interesting that in your previous posting you were all concerned that a sample must be randon. But here you are faced with a clearly non-random sample and you swallow it whole.
Also interesting that you wish to ignore the majority of professors and weather people, but you latch on to any individual professor or weather person who agrees with you. Convenient, eh?
The objections being raised here have nothing to do with the efficacy of statistical sampling.
This is the point: 97% of the authors of the sampled papers did not agree with the theory of AGW. Only one third of them agreed with it, the other two thirds neither agreed nor disagreed. How is it that the opinions of the two thirds of all scientists who do not support the theory are irrelevant?
estesbob did say “So we can make claim that 97% of scientists believe in man made global warming based on the review of 3,800 papers published over a 10 year period?”. That review of 3,800 was a statistical sample. It sure seems to me like he was objecting to the sampling process.
The “sampled papers” are the papers that took a position on global warming. It is true that 97% of those papers agreed. The fact that those sampled papers can be thought of as existing in a larger group of papers that popped out of a search has no bearing. You can just as well consider those 3,800 papers together with 10,000 papers on the works of Shakespeare, and then note that none of the papers on Shakespeare took a position on global warming. Would that in any way taint the conclusion that of the sampled authors that wrote about global warming, 97% agreed that it was real? Sure, we would like to know what the other 2/3 of the authors thought. But we have no information on that. We can’t even say that they are undecided. They could very well be decided. In fact they probably are, and in favor of global warming. Just as with any statistical sample, we confidently apply the results of the sample to the whole population. The only time this is not valid is if the mechanism for selection is biased in favor of one outcome or the other. Is there a reason for thinking the selection criterion (that your paper took a position on the reality of global warming) has any correlation, either positive or negative, with your opinion on the reality of global warming? As I said before, those 2/3 of the papers that did not take a position were just not about that, just like the Shakespeare papers.
No, this all makes sense. Imagine exit polling during an election. A news source reports “Exit polls consistently show that Joe Shmoe will win the Senate race”. This, despite the fact the the exit polls may only have asked 1% of the voters. No one would say the news source was absolutely false.
Not if it came out that the news crew first asked whether the person was a democrat, republican, or independent. And then only reported the results of the independents as representative of the total population.
You will see that I already made that point in the paragraph above the one you quoted.
The only thing the survey proves is that over a 10 year period 97% of 3,800 self selected papers claimed global warming was caused by man. There is absolutely no way you can twist these “facts” into 97% of scientists believe in man made global warming
Why do you continue to question the selection process? They are simply all the papers that took a position on global warming. If you have evidence that this was not the selection criterion then please present it, or else stop making this claim.
Correct. It is a serious error to claim that this was not a random sample, since the survey was about the population.