There are good discussions and/or examples of "confounding" in some of the statistics textbooks in current use (e.g., Rossman, Moore's BPS). Nevertheless, it doesn't seem to be easy to make this clear to students--at least not in a single encounter.
Moore & McCabe's IPS shows on p.201 a nice diagram that helps distinguish among these situations: (a) changes in x cause changes in y (b) x is correlated with y because each is affected by changes in a lurking variable z (c) the effect of x on y is unclear due to the existence of a lurking (confounding) variable z that causes changes in y I think the diagram usefully augments the words and examples we use, and I wish this diagram had not be eliminated in BPS.
Anyway, sometimes a current example can help. There was a neat one in an AP release that appeared in my local newspaper on Nov. 21; so it probably appeared in many other newspapers around the country about the same time.
Briefly, the article reminds readers that it has long been known that American blacks suffer from cardiovascular diseases at higher rates than American whites. But it now appears that the difference was not race; it was place! That is, the higher rates of American blacks are almost entirely due to higher rates among southern blacks. Norhern blacks and Caribbean blacks did not have the same high rates as southern blacks.
Unfortunately, the study was not able to compare southern whites with northern whites; so the study's results are not yet conclusive.
I think this could make an easy-to-understand classroom example. If you can't find the article in a local newspaper, my copy says that the study "was published in Thursday's issue of the New England Journal of Medicine." Since Nov. 21 was a Thursday, I supppose that could mean the Nov. 21 issue of NEJM; or maybe Nov. 14. Either way, it shouldn't be hard to find.
============================================== Bruce King Department of Mathematics and Computer Science Western Connecticut State University 181 White Street Danbury, CT 06810 (firstname.lastname@example.org)