According to most math books, the correlation coefficient is the linear association between two variables. Bottom line: correlation coefficient shows the relationship (or, association) between two things.
The correlation coefficient NEVER shows or proves causation. A USMLE high yield question may include answers with the word “cause” in it. Never select an answer-choice with the word “cause” in it.
Pictures are worth a thousand words so the USML Ebiostatistics workbook provides pictures and arrows that highlight everything you need to know about this topic.
The Correlation Coefficient is a number between, -1 and 1 (including -1 & 1), and is denoted by the letter, “r” (e.g., r = 0.78). r provides the direction & strength of a correlation between two things – the, “linear association between two variables.”
Focus on one and only one row; focus on the row of the specific illness or condition being questioned. All other rows are irrelevant AND you will ~never use the percentages (if given) to compute the answer.
Let’s say the question is, “What is the case-fatality rate for pancreatic cancer?”
If r is between -1 & -0.1, then the direction is .-1 = perfectly negative correlation (maximum strength), so a -0.6
represents a mildly-strong association.
If r is between +0.1 & +1, then the direction is .
+1 = perfectly positive correlation (maximum strength), so a +0.8
represents a strong association.
Here’s a general scale for Correlation Coefficient’s strength:
Questions may provide scatter plots (the easiest) or, entirely in the form of a word problem (trickiest). Calculations are not required; instead, it’s entirely conceptual.
A high yield question designed to confuse the examinee could provide a positive r, p=0.02, but frame the answer so the correct choice involves two decreasing variables. I provide practice questions that address this.
The USMLE biostat workbook describes this topic by using many pictures, additional details, a wide variety of word-problems with thorough explanations, and much more. Buy now!