Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
causation vs perfect correlation
Message posted by Han on November 23, 1999 at 12:00 AM (ET)
If a set of bivariate data gives a high positive correlation or even a perfect correlation of 1, and still, it doesn't imply causation, then what's the point of finding such a mathematical relationship?
In other words, what can we conclude from the result, if it can't be used to support a hypothesis?
(If a example can be used to explain this more appropriately, please do so)
Thank you so much!
READERS RESPOND:
(In chronological order. Most recent at the bottom.)
Re: causation vs perfect correlation
Message posted by Bill on November 23, 1999 at 12:00 AM (ET)
The Pearson correlation can be used to support a hypothesis. In general it tests the hypothesis that the observations in your study are from a population in which the correlation between the variables is zero. If you want to test the hypothesis that X causes Y then you will need a different experimental design and statistical analysis. The correlation coefficient may be used to identify relations (cholesterol is related to heart disease). Subsequent studies may identify the causative factor for the outcome in question. It may be that X did cause Y but that will be determined by further study. It's another tool in your investigation. The correlation may also be used for prediction. Students' ACT (or SAT) scores may predict subsequent college performance (GPA). ACT does not "cause" GPA but it may be useful in predicting college success. Thus, students with high or low probability of success may be selected or not selected for admission to the institution.
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.
