Wikipedia:Reference desk/Archives/Mathematics/2013 March 2

= March 2 =

Correlation =//= Causation... Ever?
If there is a strong positive correlation, then that is still a correlation, right? That can't be considered a causation, because it is still a correlation. My reasoning is that any correlation, no matter how high or low, can be tampered with by selecting a critical r value and comparing the critical r value to the obtained r value (Pearson product-moment correlation coefficient) in order to achieve a desirable result? For example, there may be a lot of evidence (very strong correlation) to suggest a statistically significant change, but because the statistician chooses a ridiculously high critical value, the statistician fails to reject the null hypothesis and maintains that there is no change, because there is a conflict of interest in the experiment that may influence the statistician's own life in some bad way, which the statistician obviously denies in order to preserve his academic credentials even though lying is liable to being accused of academic dishonesty and thus can greatly ruin his career. Are there real examples of such things happening where statisticians intentionally manipulate the data in order to achieve a desirable result and report the result to a scientifically and mathematically illiterate general public? Sneazy (talk) 06:39, 2 March 2013 (UTC)


 * I can't make sense of all that, but there is a famous book called How to Lie with Statistics that is probably the sort of thing you are looking for. Looie496 (talk) 06:47, 2 March 2013 (UTC)


 * Actually, the first sentence is the question. Sneazy (talk) 06:55, 2 March 2013 (UTC)


 * Actually there are two questions and one statement with an appending question mark. Looie answered your second question. As for the first, indeed correlation is correlation and not causation. Your reasoning is somewhat convoluted and the agency of a statistician need not be invoked to show that they are fundamentally and conceptually different. — Preceding unsigned comment added by 202.65.245.7 (talk) 08:51, 2 March 2013 (UTC)


 * The article on Statistical hypothesis testing explains how hypothesis testing is normally done. It's not just comparing r-values, though a high r-value suggests that there is something to investigate (which might or might not be a causal link).  I assume you've read the article Correlation does not imply causation.   D b f i r s   08:59, 2 March 2013 (UTC)


 * In the real world, we assume causation all the time, when all we can really prove is correlation. For example, observing the correlation between smoking and lung cancer deaths was taken as evidence that smoking causes lung cancer deaths, before the specific mechanisms were discovered.  It might have turned out that people with a certain gene which causes lung cancer also, as a result of that gene, craved cigarettes, but this seems unlikely.  StuRat (talk) 18:00, 2 March 2013 (UTC)


 * ... but causation was widely denied (especially by tobacco companies and smokers) even when the correlation was shown to be strong.   D b f i r s   08:24, 3 March 2013 (UTC)


 * Yes, and we're still with the problem that no specific case of lung cancer can be blamed on smoking, which is a problem for those wishing to file lawsuits (of course, one could also argue that anyone who started smoking in the last few decades had fair warning of the danger, but that's another issue). StuRat (talk) 16:26, 3 March 2013 (UTC)

Restricted polyabolos
Does anyone know the number of possible restricted polyabolos?? Please see Talk:Polyabolo for what this means. Georgia guy (talk) 20:41, 2 March 2013 (UTC)


 * The sequence begins 1, 2, 2, 6, 8, 22, 42, 112, 252, 650, 1584, 4091, 10369, 26938, 69651, 182116, 476272, 1253067, 3302187, 8733551… . —Bkell (talk) 01:02, 3 March 2013 (UTC)