I'm not going to bother with a point-by-point rebuttal of what's been said recently. I wanted to hit a few points.
http://www.biotopics.co.uk/newgcse/measuring_population_sizes.html
Here is a discussion of methods for determining the size of a population based on sampling. These methods are extensively used in biology, but they are not limited to biology; the math works regardless of whether you're dealing with squirrels, beads, or scientific papers (I've checked all three). It's not perfect; these methods give an estimate. However, they will let you know if you've found most of the frauds, a few of them, or hardly any of them.
Thus, we CAN know--or at least formulate a working hypothesis for--the number of frauds in a given dataset.
One method not discussed in that link is rarefaction.
Here's a wiki article discussing it. This provides an additional test, in that it allows us to know if we've adequately sampled the frauds or not. These methods can again be applied to non-living things. There are some issues with rarefaction, but again, it will give us a good working hypothesis for whether we've identified most, some, or few of the frauds in a set of scientific papers.
After that, in this thread we have one datum and one argument. Everything else has been mere speculation and bickering (note that I'm tossing my own reference and arguments out here--I'm giving you guys the best of it):
Datum: Known frauds in a particular dataset compose less than 1% of 1%.
Argument: We don't know the percentage of unknown frauds.
From these, there are two defensible positions one can take:
1) Tentatively accept the datum as valid, and conclude that scientific fraud is rare. Note that it's a TENTATIVE acceptance--once more data are provided this stance should be re-examined to determine if it still holds true. There is data to support this conclusion, however, so it's not untennable.
2) There's not enough data to draw any conclusions. This is defensible, IF one stops at that point. I'd say that if one wishes to continue in this discussion but holds such a view, one is obliged to attempt to gather enough data to draw conclusions; thus my providing a viable methodology.
What is unacceptable and irrational is to take the datum we have and the argument we have and say that since we don't know the rates of fraud, we therefore can conclude _____. ANY statement that fills in that blank is unsubstantiated.
If one takes the datum and the argument and concludes that science is hopelessly flawed and full of frauds, one has completely abandoned any pretense at rationality. There is no data to support that conclusion, and it directly contradicts the data we DO have.
What is also unacceptable is to use data from other fields to draw conclusions about science. It seems reasonable, and it would be under certain conditions, but the present argument does not present those conditions. This type of argument uses what's known as a proxy character, and is something I'm extremely familiar with. We use them all the time in historical sciences; often data cannot be tested for directly, so we find ways to come at it sort of sideways. Dental wear in predators is a proxy character for predator population size in an ecosystem, for example. Certain osteological traits indicate prevelance of chairs. Ritualistic burial indicates abstract thought. That sort of thing. The issue is, there are always confounding factors that must be addressed, and these can only be used in cases where no actual data about the topic is available. To use proxy characters when data on the actual topic is available is to say that the actual data doesn't matter, which is always wrong in science. Proxy characters are, by their nature, very low quality data--they are Argument by Analogy, which always leaves open the posibility that the analogy is flawed.
In science you want to focus on the highest quality data--meaning here the data with the fewest assumptions and confounding factors. Basically, when you have data on the topic, you use that data. When you don't have data but can obtain it, you obtain that data. Proxy characters and Argument from Analogy are only acceptable when you do not have direct data
and cannot obtain it, and even then it's not irrational for people to reject the proxy data, since it's not actually about the topic at hand. The bolded part is why I reject the notion of using fraud rates in other areas to study fraud rates in science. It's lazy. That's not to say that the data are worthless--they can indicate areas to examine, certainly. Studying fraud in other fields can demonstrate how it is perpetrated and provide us with methods for detection. But when one is presented with actual data specifically on the topic at hand, and proxy characters, one cannot rationally discard the actual data in favor of the proxy characters. Since we have actual data it is incumbant upon us to use that data, or to analyze the actual dataset to demonstrate flaws in it (such analysis will inevitably provide better data, so the proxy data is irrelevant).
In conclusion: We CAN estimate the rates of undetected frauds. Therefore if someone believes that the rate of detection is significantly lower than the rate of commission, they are making a testable claim. This is a science forum, and methods for testing that claim have been provided. I invite anyone who claims that fraud rates are much higher than detection rates to demonstrate the point. If you think that fraud rates in other areas demonstrate significantly higher rates of fraud in science than have been presented, it is incumbant upon you to provide sufficient data to demonstrate that conclusion. If you think that science is insufficiently inspected or regulated, it is incumbant upon you to demonstrate that this is the case--and that means demonstrating high fraud rates--and that means, given the data available in this thread, demonstrating that perpetration rates are significantly higher than detection rates.
justintime said:
If you have been following recent criminal trials in the US. There has been many acquittals where the lay jury ignored forensic evidence and science because it sounded like a load of esoteric high flung dung.
This has nothing to do with the scientists. This is the fault of the lawyers.
If scientists cannot even communicate their trade, it is because their profession does not develop communication skills and is designed to deliberately mislead, mask, evade with over the top mambo jumbo high flung dung, esoteric for a purpose....ie. to avoid public debate, discussion.
Nope. A much better explanation is that scientists generally speak to other scientists. We develop tools for very effective communication with the audience we are generally dealing with. Those tools do not necessarily work with other audiences. A teen pop star isn't going to be popular among the opera-going crowd, but that doesn't mean the pop star is a fraud.