Personally I'm more than willing to get worked up. I've had debates where I let my lesser nature come to the fore. You aren't going to offend me in any way. I would be the world's largest hypocrite if I were to complain about "snarkiness".
Fantastic, you and I will get along famously. Like you, I would be a hypocrite to complain about agressiveness or some such wimpy complaint. I liked what I heard about Luis Alvarez, a nobel physicist.
"Luis Alvarez died on September 1, 1988, ending one of the most versatile, successful, and combative careers in modern science. At Snowbird II, held just six weeks later, one of the participants proposed two minutes of silence in his honor. Walter rose to say, "My father would have been mortified. He'd much rather have a good fight in his memory.'" Lawrence Powell, Night comes to the Cretaceous, (New York: Harcourt Brace and Co., 1998), p. 165
I hope my children can say the same thing about me!
Also be mindful of the statistics not just the "raw data". Noise and variance components both have an impact. Statistics allows us to cut through the natural variation in data.
but, as you know statistics only works if one has the right model. Is 1,2,2,1,1,1,2,2,2,1,2,1,1,2,2,1,1,1,2,1,2,1,1,1,
a random sequence?
If it is a coin, it might be (I didn't run the test on it). But if it is a six-sided die, it is decidedly non-random. It wouldn't be if it were the output of the program statement A=rand(100), which would allow a variation from 1 to 100. Statistics only works well when one has the right model of the thing one desires to analyze statistically. My point with the linearization of the data, that is the standard tool of global warming even if it is a hammer used to saw down a tree.
Gauges get ignored or, if not ignored, then badly set. The robustness of the overall trends is something that can be determined. In the present case of global temperature we are demonstrably not limited to solely unit-by-unit land-based temperature gauges. The fact that two unrelated guages show similar trends is usually an indication that neither gauge is horridly off.
I agree that gauges get ignored, and they get set wrong. But, when one becomes aware of it, the professional thing to do is to fix it, not collect more bad data. In Watersville Washington, they decided to ignore and keep collecting bad data. That isn't professional. They are paid to know that their temperatures are correct. They are failing that part of their work.
In addition, significant amounts of effort are, as shown earlier in the IPCC report, taken to assess bias, correct for bias and, on a broader scale, compensate for individual guage biases.
I will tell you, most scientists in China wouldn't trust historical empirical data in China. During Mao's era everything was political. If the boss said July was the coldest month, it would become the coldest month. This skepticism applies to much of the data in China. You can't even trust the oil well records. Years ago, I believe it was Exxon who bought a big lease onshore only to find that the records didn't tell them that the structure they bought had already been drilled years earlier. Empirical data out ofChina esp. during Mao's era is questionable. So, lets look at some of the Chinese data, which IS used in the calculation of global temperature and used as if it was as good as the bad data in the US.
Below is a picture of two chinese stations. I downloaded this data from the Chinese Government site.
http://bcc.cma.gov.cn/Website/index.php?ChannelID=43
Unfortunately in the readme file, the Chinese names don't come across. While I read some Chinese, I can't decode the ascii of Chinese names. So, I used station number in this data. The station below is from 2 towns only 52 miles apart. These towns are in SW china, in Yunan province. That I can tell from the lat long. I have been there (a beautiful part of China) but one shouldn't get a 24 deg C difference in annual average temperature over short distances, even in that part of the world. That is 41 deg F difference. this data is crap, even crappier than that obtained in the US. Yet, it goes into the global warming calculation and everyone treats it as if it is as good as our data.
Some people here may not want to look at the data, but I find that utterly incomprehensible when this is what the data looks like and said data is used to tell us that the world has warmed. I don't think it is good enough.
It is a great thing for the one group to go around and rate the temperature systems on an individual basis. Indeed it will surely help generate better data in the future. But that is like look at one set of gauges while ignoring how others "overlap it".
I am all for a good survey. But, it wasn't our government that felt the need to take pictures of their sloppy work.
Actually ignoring statistics is probably a worse crime. Since data contains noise, statistics helps quantify noise in an effort to get at the best answer.
I've seen perfectly smart scientists make decisions on graphs without statistical analyses behind the data points. I count that as worse than useless because there is no idea where on the map you are with just a single or small handful of data points.
You ought to try getting scientists to understand risk and probability. That is how we in the oil business make money--by assessing risk correctly. If we do, then we know how to invest. But many don't seem to be able to handle chance of success very well.
There's two problems to face: Type I errors (erroneously rejecting a true null hypothesis: "false positive") and Type II errors (erroneously accepting a false null). We can never ever remove both. In order to eliminate one you have to accept you increase the chance for the other.
In the case of global warming the null is reasonably stated as "There is no (anthropogenic) global warming trend in the data". I am merely testing the data to see if I can reasonably avoid making an error in rejecting that null (which is, effectively, the "Climate Skeptic Stance".)
I won't say that there is no warming. I think there is some evidence that parts of the earth have warmed. I don't think the land data is sufficient to know how much it has warmed. But, the question is whether or not the northern hemispheric warming is simply part of the natural phase of it warming when the south cools?
We need to start talking about the urban heat island effect. Maybe tomorrow. I usuallly work my ranch on the week ends and have no internet up there (or TV). So, I might post some stuff tomorrow but it may be sunday before I post again.
The p-values have become very important for me when assessing the data and possible correlations. It keeps me from merely relying on my "view" of the data plots.
But, if the data upon which you depend is so bad, can the p-values really mean anything? Especially data from the likes of the chinese example below.
(Again, I am still relatively new to the world of statistics, so I hope I have not misstated a point here, hopefully a statistically savvy poster will correct me if I'm mistaken).
p values are very very important. but one still must have the correct model. modeling the out put of a coin flip with a six sided die gives you a different p-value than modeling it with a two-state system.
One of my papers With Simons and Yao, discussed the inability of the current statistical tools to model large DNA sets. Very slight deviations from the model made for bad p-tests. In other words, when you have a string 3 billion characters long (as we did with the human genome), the genome becomes its own model. Any deviation from that genome would fail the test. That is the Journal of Statistics and planning paper.
http://www.sciencedirect.com/scienc...serid=10&md5=2536cd9788d5b9293702b8d7ba943456
