"The same is true with the correlation coefficient (and statistics in general). The formula for a correlation coefficient is like a recipe. The information that goes into it (∑X, ∑Y, etc.) are the ingredients. If you put the wrong ingredients into the formula, you will get a number that may very well look like a correlation coefficient. The only problem is that it is a defective correlation coefficient. Therefore, it will not mean what it should mean."
Feel free to be the first
Please review the linked page for context.
If you can think of something better than this,
please add it to the database