Another shooting in the USA. Remind me about the reason for having guns again. - by SubJeff
jay pettitt on 7/9/2012 at 09:54
Quote Posted by "Chimpy"
So I thought I'd convert to total murders and churn out some scatter plots.
So unless there's some other reason for the high homicide rate in the US compared to its peers, it's not obvious that US gun ownership is discouraging total murders/100k.
DDL on 7/9/2012 at 10:22
Quote Posted by Subjective Effect
Certainly. But is the causation the fact of having an eye patch in and of itself (which would clearly seem ridiculous) or something else related to it?
The problem here is you've deliberately picked a ludicrous example for a putative causation hypothesis, and are now using the exact same example to demonstrate the ludicrousness of any possible causation.
You're saying "check out my silly thought experiment!" and then when anyone actually points out flaws, you respond with "BUT LOOK ITS SO SILLY"
If we just kept it generic (i.e. two otherwise indentical individuals have a single distinguishing characteristic, and one is statistically more likely to die from a particular cause of death), then everything Al_B says is entirely valid: if the numbers back it up, you'd have good reason to investigate further. It doesn't really matter what the characteristic
is, if the numbers back it up you have a case for a possible causation.
EDIT: also, thanks for fix0ring my images, Al_B!
SubJeff on 7/9/2012 at 10:31
Yes, you have a case for a possible causation. I never denied that.
I only gave that example to illustrate the difference between cause and correlation. I'd already tried to explain it several posts ago which why I've resorted to using clear examples of where correlation is not the same as cause.
DDL on 7/9/2012 at 10:41
Yes, but in your initial post on this matter you stated "None of this matters. Correlation is not causation."
While the second statement is true, it's not the whole story: correlation is associated with causation (it correlates, in fact :p), correlation can imply causation, it's simply not exactly the same thing as causation.
"Smoke isn't fire": true, but if you find the former, worrying about the latter is generally prudent.
More importantly, the second statement does not in any way make the first statement valid.
A more thorough way to address the correlation/causation matter is to suggest a causative hypothesis, then look for correlation that supports or invalidates that hypothesis. This is exactly what people are doing here, and that is EXACTLY why 'none of this matters' is a wholly incorrect statement.
SubJeff on 7/9/2012 at 10:55
Prior to Larry Gs graph being posted we'd already mentioned about other reasons for disregarding this lack of correlation though.
jay pettitt on 7/9/2012 at 11:20
At some point, you've got to ask yourself - what are the odds that all this is random or chance?
With the stuff we're looking at here, you wouldn't want to draw many conclusions. But if you work hard to dismiss alternatives and if supporting evidence comes along then those odds get smaller.
LarryG on 7/9/2012 at 14:10
Quote Posted by Subjective Effect
Correlation is not causation.
True. Causation is one form of correlation. There are other forms of correlation. Correlation is the weaker relationship. If you don't have correlation you cannot have causation. So a failed correlation test is sufficient to demonstrate that there is no causation.
SubJeff on 7/9/2012 at 14:20
No it isn't. It's sufficient to show it in specific cases.
We can say that there is no correlation between percent of people owning guns and gun crime per population, if we compare different countries.
We cannot say that a lower percent of gun ownership in a specific country will not reduce the amount of gun crime per population because of the specific laws, gun culture,type and rate of criminality which differ in different countries.
And causation is not a "form" of correlation. They are different concepts.
jay pettitt on 7/9/2012 at 15:46
I think SubJeff's point is that you can refine your questions and your data. So the UN data for the globe doesn't show a correlation for gun ownership/homicides but a more refined comparison of Countries grouped for similar economic and social metrics might show that ownership and homicides are correlated.
If you fail, it's just for that data and that comparison.
