Science: Chaos and a piece of rope, Free Energy(?), Maxwell's Deamon, New periodic table, Nu periodic table, Nu periodic table, Occram's razor, Poetry in mathematics, Prince Rupert drops, Rules of science, Science of conjecture, String Theory, Directory
******
Much is made about errors that are by either humans or machines but nothing is said about the role that statistical variation can play in providing false or misleading interpretations of data. Sadly, few scientist have an understanding of statistics. Too often, the conclusion is; it happened or it didn't happen. That's wrong. When you begin to subject data to a probability analysis errors in judgement, are reduced to a science. Let me give an example from personal experience:
While at the National Institutes of Health, in Dr. Harry Gelboin's lab in the Cancer Institute we were pursuing the idea that certain chemicals would cause a reading of a cell's DNA library resulting in a particular messenger RNA being formed and that specific proteins would be synthesized accordingly. We hypothesized that this would be an orderly process with a lag phase following application of the chemical, induction of RNA synthesis, over-production of RNA, reduction from the high level of synthesis to a level below the resting state and finally an increase to return to homeostasis. The experiment was simple, measure activity in a liver homogenate as determined by incorporation amount of radioactive nucleotides, administer the compound (phenobarbital in our case), and at time intervals, sacrifice animals and measure(1) the level of incorporated nucleotide. We did the experiment and we got a beautiful sine wave just as predicted.
Great! However, we never could reproduce these results. Convinced as we were, we just couldn't get Mother Nature to smile again. And here is where statistics comes into play. At each time point, a group of observations was made and each had a mean and variance. Not only were multiple animals, but multiple assays done as well. The old (n-1) provides some help but not much, this just means you have to have more than one "measurement." It's the power of numbers that finally must be the answer - either a lot of observations or the numbers be tightly packed with a small standard deviation. And that's just how the experiment was conducted. And yet the following experiments with identical conditions and number of animals, never gave the sine wave.
In the case of phenobarbital induction of RNA synthesis, reluctantly, we had to conclude the answer was no, there was no ordered and systematic progression of events. Mother Nature had actually slipped us a sticky wicket and if we had bit, and rushed to publish the one experiment prior to making the repeats - disaster! But we didn't so the data remains safely in the lab notebooks.
Such it is in the case of data dredging today. Researchers (or lawyers) look for instances where some event occurs and try to relate it to an environmental cause. Of course if you look hard enough and refuse to accept data that is contrary to your precept, you can put together a rather impressive case of cause and effect. Examples are everywhere; power line emissions, ground water pollution by any number of substances, exposure to asbestos, smoke, diesel engine emissions, exposure to the sun, mercury in seafood &c. the list is endless.
Does Mother Nature tell the truth all the time? Or, is She just trying to teach us some of the elementary laws of numbers and statistics.
***** (1) Note that there is a major difference between measure and determine. You measure in an attempt to determine.
****
Joe Wortham's Home Page, About Joe Wortham, Directory of Web Pages
Questions? Comments? [email protected]