The more I know about science the more I know scientists don't do it. And usually it isn't that bad of a thing, if we go down the wrong path for a couple of years it's no big deal. However, when there is a political agenda behind the science, the the scientists themselves are in a position where if they find out they are wrong they are out of a job we have a really good combination for really bad science. Can anybody guess which science I am talking about? That's right, the one that is put in charge of proving that man is causing global warming. This is the same one that predicted that there would be significantly more hurricanes last year, and that 2007 would be the hottest on record.
I don't really think the scientists are being devious when they make their conclusions, but I do think that they are selectively ignoring some facts while using shaky statistics to prove their ideas. A really good example of a guy that knows enough statistics to get himself into trouble is this guy. My main problems with his analysis is that he finds out which lag to use by the one that fits the model the best, and then goes on to say that after changing the data to the best lag the data fits the model really well. Well duh. 
Also if you look at his final projection of the temperature it is quite within the bounds of the prediction that we will always be cooler (or at least the same temperature) as right now, i.e. the best model still does not predict a significant raise in temperature.I wish it was just this one guy, but the more I learn about statistics the more I see how they are just using it to prove their point, not to discover the truth.
5 comments:
That's why the only statistics I pay attention to are baseball statistics.
Except for fielding stats. C'mon, those guys are making most of that shit up!
See, I never found a use for statistics. My hypotheses were always wrong. Stupid reality always got in the way of nice, simple math.
And that's why the scientists make most of it up as they go along.
Aw, anybody can use statistics to prove a point, 38% of the people know that.
I ask my friends, hey, how come the only people I know who aren't convinced Man is causing global warming are engineers? The answer: Engineers know how to analyze data. Well, scientists too, but as you suggest, the need for grant money skews their results.
First of all I find your first sentence obnoxious. Are there political agendas that influence science? Yes. Are there people who let their own opinions influence their conclusions? Yes. Does that mean that science can't be trusted or is not valid in general? Absolutely not!
Now onto to your complaint about predictions. You should know as well as anyone that predictions are not always going to come true. They are the best guesses of what is likely to happen, but that doesn't mean they will be perfect. Your weather man may predict snow in your area but it may in fact not snow. Your stock broker may predict that Microsoft stock will take a plunge but then it may not happen.
All right let's move on to Dr. Loren Cobb who you criticize as "knowing just enough statistics to get him into trouble". God those crazy scientists that know nothing about statistics... I mean it's not like he got his PhD in mathematical modeling or taught statistics for 15 years... (http://www.ndu.edu/chds/journal/papers_bios/LorenCobbPhD-bio.htm)
I mean geez what would he know about stats? It's not like he's had several papers published in the Journal of the American Statistical Association.
As far as your specific criticism of his statistics, I don't really get the problem. If I recall from my mixed modeling statistics class (taught by a "real" statistician not those crazy science people), when you want to test the prediction accuracy of two models, don't you first make sure the
models are both optimal (removing factors that are not good predictors and so on)? Let's imagine you are testing how well two pesticides kill bugs. Now you can imagine that these two pesticides could have different lag times for killing the bugs. Wouldn't you first find the optimal lags for each model and then test which model best fits the bugs death? Now I am not a statistician so I may be wrong. I'm also not sure how you are judging the "significant change in temperature"
From the graph, it could have a decrease or a similar temperature or it could be dramatically higher. I think the important thing to notice is that the overall trend and the best guess for what will happen (regression line) is that there will be an overall increase in temperature.
You talk about global warming as though it's some kind of government conspiracy. First of all, the government has rarely done anything to try to fix this problem that all of the scientists have been screaming about. If anything, I would suspect them of covering up any evidence of global warming to keep their pockets well lined with oil money. If it were one or two scientists who were arguing that this is a big problem and the rest were not then I could say yes they are just trying to keep their job and get more grant money by making stuff up (think Cold Fusion scientists here), but when you have 2500 scientists from all different countries finding consensus on the occurrence of global warming and on the human factor in causing global warming, it seems pretty hard to ignore. Certainly, you can try and many people do. There are certainly a lot of Christians out there who plug their ears and close their eyes to all of the evidence for evolution. Science is wonderful because it is self-correcting. If the evidence doesn't fit, someone will find it and correct it. When the evidence continues to fit, as it does in global warming... where out of 928 published papers in refereed scientific journals between 1993-2003 75% of them indicated that human factors were significantly contributing to global warming and the
other 25% were looking at trends in ancient times (before industrialization) and made no hypothesis about the current contributors to global warming, and 0% reported that human factors were not significant contributors to global warming.
If you'd like to read a bunch of scientists report on global warming, try this page... http://ipcc-wg1.ucar.edu/wg1/Report/AR4WG1_Print_Ch09.pdf You might find pages 702-703 particularly relevant.
This is another nice site about the consensus of global climate change.
http://www.ucsusa.org/ssi/climate_change/climate-consensus.html
Finally, lets think about our options in terms of hypothesis testing and errors.
Hit (reject null hypothesis when null hypothesis is false)
Yea we saved the world from it's ultimate demise by stopping green house gasses that would have destroyed the planet.
Correct Acceptance (fail to reject null hypothesis when in fact null hypothesis is true)
Yea we didn't do anything about green house gasses and we didn't all die or waste any money.
Type I error (False alarm- reject null hypothesis when in fact null is true)
Damn... we spent a lot of money on technology and reducing green house gasses for nothing. Maybe we produced a temporary recession (maybe).
Oh look we had some friendly side effects-- cleaner air locally by decreasing overall polution and also possibly some money saved by replacing fossil fuels produced in foreign lands with domestically produced renewable energy.
Type II eror (Miss- fail to reject null when it is in fact false)
Oh really damn. We didn't stop green house gasses and now we've screwed the world over. All the animals and plants are dying.
Now certainly I may have exaggerated it some, but it certainly seems to me that making a Type II error on this topic is a hell of a lot worse than making a Type I error.
Can people lie with statistics? Yes. Do scientists (and statisticians) do it? Yes. My advisor has a fabulous quote on his desk "The data will tell you anything if you torture them long enough." Do I think that this man misused statistics? Not from my quick look, but I'd have to see his model in more depth (like his or others actual published scientific papers instead of the "Quaker Economist" version which seems to have simplified things a little). I may even consult with another statistician whose knowledge of modeling would be vastly superior to my own basic knowledge. Does any of this mean global warming isn't occurring or doesn't have a significant human factor? I doubt it. 2500 scientists seem pretty hard to argue with. I'm not saying they couldn't be wrong, but I think the changes are small (p<.05). Am I willing to risk a Type II error? Nope!
Post a Comment