Um... you're looking for a correlation coefficient, not a p-value. P-value is say that a result is outside the expected value (such as saying a coin flips heads 5 times in a row -- the p-value of that result is very low, assuming a fair coin). Correlation coefficients are comparing the results of two sets of data, and saying if there is a statistically significant correlation between them. But then, correlation is not causation -- for example, ice cream sales go up with local temperature (people buy more ice cream on hot days). But buying more ice cream will not make the weather warmer.
This was the very first thing my psychology professor taught us in PSYC 101.
ReplyDeleteIt's also the first thing many scientists seem to forget when his pet theory get's a low p value.
DeleteUm... you're looking for a correlation coefficient, not a p-value. P-value is say that a result is outside the expected value (such as saying a coin flips heads 5 times in a row -- the p-value of that result is very low, assuming a fair coin). Correlation coefficients are comparing the results of two sets of data, and saying if there is a statistically significant correlation between them. But then, correlation is not causation -- for example, ice cream sales go up with local temperature (people buy more ice cream on hot days). But buying more ice cream will not make the weather warmer.
ReplyDeleteWell yes of course. But I don't think it's a random coincidence that ice cream sales go up on hot days like in the examples above.
DeleteMy question is: "Why?? did someone choose to look at these correlations in the first place?".
ReplyDelete