LAUSR

Assessing the likelihood that purportedly random samples are indeed random [1] is a powerful tool for diminishing the problem realised by Ioannidis, that ‘false findings may be the majority or even the vast majority of published research claims’ [2]. Concern over results that are too good to be true was emphasised when Ronald Fisher calculated the probability that Gregor Mendel tidied up his data to an extreme degree: ‘The data of most, if not all, of the experiments have been falsified so as to agree closely with Mendel’s expectations’ [3]. Despite breaking the rules of empiricism, Mendel’s central hypo thesis – that most variation observed across generations is due to recurring frequencies of characteristics that are not genuinely new – was true. This raises the false argument that we should be wary of throwing the baby out with the bath water, i.e., that we should not be overly concerned about conclusions bolstered by p values that are too high or too low [4]. That is a false argument, both on its face and because the noise in systems can be more telling than the clear signal [5]. Mendel’s results were so ‘clean’ that they obscured evidence of mutations, which are the building blocks of evolution. Although it may not have been Mendel’s intention, his tooclean results impeded acceptance of the evolution by natural selection on the basis that there was nothing really new under the sun – nothing for natural selection to select – only repeats of many different combinations. Contrary to Wilson’s surmise [4], Carlisle did not overstate his case [1]. P values have no value in reference to non-random samples [5] and truncating variance to increase statistical significance, or lack thereof, should not be allowed to worm its way into acceptability.