Talk:Spurious relationship

Untitled

I am appalled that this page has been listed on votes for deletion. It is a very nicely written -- even if stubby -- article on a topic of importance in scientific inference. Michael Hardy 01:35, 29 Feb 2004 (UTC)

Likewise, but not surprised, anymore. -- Cimon Avaro on a pogostick 01:56, Feb 29, 2004 (UTC)

Guys, don't you know in America the word "relationship" means something else? Doubtless, a relationship based only on ice cream is spurious, and hence the article was a tautology. Mikkalai 02:52, 29 Feb 2004 (UTC)

What is the difference between spurious variables and antecedent variables? I can't tell from the article. Robert Southworth 19:27, 29 Feb 2004 (UTC)

I can't tell either. I'm editing the article appropriately. --DAD 01:29, 19 Dec 2004 (UTC)

"Confounding variable" is redirected to this article, but "confounding factor" has its own stub article that contains essentially only the ice cream example and links to two other articles, one of which is this. Shouldn't confounding factor be redirected here also, and that stub be deleted? -- 65.24.92.175 14:47, 15 Jun 2005 (UTC)

• I say no: I came to this page looking for an explanation of confounding variables. Thankfully, I found the link to the confounding factor page, which gave me what I needed. Confounding variables are an important topic in experimental design, and simply redirecting the page here is misleading. I've redirected confounding variable to confounding factor, added the stats-stub tag to confounding factor, and added a mention of experimental design to it. -Kieran 09:26, 21 October 2005 (UTC)

Error in this Article

Hey, this isn't right, could someone fix it? From the article:

> In practice, three conditions must be met in order to conclude that X causes Y, directly or indirectly: > * X must precede Y > * Y must not occur when X does not occur > * Y must occur whenever X occurs

> Spurious relationships can often be identified by considering whether any of these three conditions have been violated.

> The final condition may be relaxed in the case of indirect causation.

Now rain (X) may cause me to stay in (Y); that is, I stay in whenever it rains, but I may stay in when it doesn't rain (because it may be snowing, for example). So clearly there are obvious cases when the second condition may be relaxed, rather than the third.

In fact, the example of the pistol shot causing the death of the man doesn't even work, as the guy could die from a metiorite, thus violating the second condition, or he could be kept alive by the doctor, thus violating the third.

The only condition that seems to be really relevant is the first.

So, what was actually meant here??

I agree. "Y must not occur when X does not occur" is wrong. It's not necessary for Y to have only one possible cause. --75.49.223.247 (talk) 04:08, 23 June 2008 (UTC)

I added an exception to the second condition covering this objection. I also added a second exception to the third condition and a note that the the conditions are necessary to prove causation but not sufficient. I may have gone a bit off-topic on the latter. MacMog (talk) 00:53, 3 March 2011 (UTC)

Further on error

Again, referring to the following: > In practice, three conditions must be met in order to conclude that X causes Y, directly or indirectly: > * X must precede Y > * Y must not occur when X does not occur > * Y must occur whenever X occurs

The second condition indicates that X is a necessary condition for Y. Then it is "relaxed", indicating that it is not necessary that X be a necessary condition for Y.

The third condition indicates that X is a sufficient condition for Y. It is also relaxed, so it is not necessary that X be a sufficient condition for Y.

Finally the article says that the conditions above are necessary (not true, due to the relaxing) but not sufficient. This section was a nice thought, but its claim to define causality falls flat. — Preceding unsigned comment added by Arual kiram (talk • contribs) 16:06, 27 April 2011 (UTC)

You're right -- I'll work on it. Thanks! Duoduoduo (talk) 17:10, 27 April 2011 (UTC)

Combine "Spurious Relationship" and "Joint Effect"

I would agree that this should be combined with "Spurious Relationship", as this page only recites a different version (with a few extra comments) on the ice cream example mentioned in "Spurious Relationship: General Example". It might also work to change "General Example" to "Joint Effect", and insert the text from "Joint Effect" in its place.

Retrieved from "http://en.wikipedia.org/wiki/Talk:Joint_effect"

I took the step of deleting the redundant text in "joint effect" and redirecting to this entry. - Aug 28 2006

I removed the 'mergewith' tag, now the merge is complete. Terraxos 21:51, 2 September 2006 (UTC)

Spurious correlation

Should this stub be merged here? A correlecation is a subtype of a relationship, I'd think.-- _{Piotr Konieczny aka Prokonsul Piotrus | talk}  17:25, 27 September 2006 (UTC)

I caution merging these two articles. "Spurious relationship" is a specific term. I added the psychology aspect because when i was studying it i couldn't find it on the website. If they are merged i would want a search to identify the new page. as it did not for me —Preceding unsigned comment added by Harris77 (talk • contribs)

Seconded, forcefully, as for the merge. To me it seems like the combined article from two or even three articles should show under spurious relationship. Confounding is obviously a different thing altogether, because it's not spurious but about a misattributed causality. Still, there are now multiple different articles which basically refer to not just a causality that is spurious, but even a correlation that is taken to be for real and to suggest something more, while being nothing beyond spurious.

I'd for one like to see these articles combined under the most well-known and/or descriptive name. I'd also like to see a concise mathematical explanation mentioned in the head/introduction. I'd actually like to see the same thing done with all of the articles on fallacies, be they informal or not. But then, at least here. Decoy (talk) 07:14, 20 November 2011 (UTC)

Confuses *confounding* with *spurious regression*

This article appears to confuse *confounding* with *spurious regression*. Spurious regression can arise when time series variables are regressed against each other, even when there i sno confounding. (For example, regressing two statistically independent random walks against each other can produce "statistically significant" correlations, simply because such series have trends.) See http://www.econ.ku.dk/metrics/Econometrics2_05_II/Slides/10_cointegration_2pp.pdf. —Preceding unsigned comment added by 172.190.11.4 (talk) 17:15, 4 February 2011 (UTC)

Unless "spurious relationship" is a specific term like "spurious correlation", I think it's appropriate to cover them in the same article. Both are problems that arise with regressions, and you can say that the "spurious relationship" in "spurious regression" is the relationship between A and B. That C is a confounding variable is somewhat irrelevant, because it's the perceived relationship between A and B that is problematic. This should, perhaps, be clarified in the opening text. Fauxsoup (talk) 05:14, 27 June 2015 (UTC)

Incorrect interpretation of Höfer's paper

The article says that Höfer found that "the number of clinical deliveries was inversely linked with the rise in stork population". This is not accurate. Höfer found "there is no correlation between deliveries in hospital buildings (clinical deliveries)". No correlation is not the same as an inverse correlation. I have corrected the article. Kmasters0 (talk) 12:06, 5 March 2013 (UTC)