quote:
Dr. Cresswell: And, at least 400 such results have been published. Where else did the numbers in the analysis presented come from if not published literature?
Reply: I think you are using faulty logic. It is clear that many discordant ages are not published in the literature. The fact that one man has cited 400 does not falsify that concept, but only exponentiates the total possible number of discordances (both published and unpublished).
I'm not sure of the basis for your claim that "it is clear that many discordant ages are not published". Do you have a whole list of scientists who have said "oh, and we had an additional collection of dates that we didn't like, so they're unpublished"?
quote:
Dr. Cresswell: Besides, a relatively simple analysis of the literature would reveal if there was any significant under-reporting of outlying results. Collect as many reported results as possible, it'll take a while going through the literature but neither impossible nor expensive (a moderately capable scientist with access to on-line journals and some spare time could do it), and for each method and time frame (assume initially that different methods and ages will result in different distributions) determine the difference between each measurement and the assume aged of the sample measured. Plot a histogram of these differences and one would expect something not too far from a normal distribution; a significant step down at greater than 3 or 4 standard deviations from the mean would indicate that a lot of outliers are not reported.
Reply: This is an excellent idea. And I pondered, on more than one occasion, in doing so. But again, the concept of unpublished discordances renders such a study incapable of accurately reflecting the true scatter of ages.
What I'm proposing isn't an assessment of the accuracy of the measurement of the true age. What I'm proposing is an assessment of the number of unreported discordant results. With measurements using one technique of rocks of a supposed similar age one would expect a smooth bell-like curve (a range in the actual dates would tend to flatten the top); the bit you'd be interested in are the tails of the distribution above and below the main part of the curve. If these follow a smooth decline with decreasing number of results at greater/lesser ages then the chances of discordant results being unreported is low, if there is a sudden step either in the distribution itself or a differential of that distribution then a significant number of discordant results have not been reported.
Is that clearer?
™ Version 4.2