DQ 1 : Explain why correlation is a common statistic to measure a generally linear relationship between two variables.

The prompt asks why correlation is common, so the answer has to be about its properties rather than its formula. Pearson's r is standardised: it divides the covariance of two variables by the product of their standard deviations, which strips out the units entirely. That single feature is why it is everywhere. Covariance between body mass index and systolic blood pressure is expressed in units of kilograms per metre squared times millimetres of mercury, which nobody can interpret; the correlation is a number between minus one and one that a reader can compare directly against the correlation between income and life expectancy. It is scale-free, symmetric, bounded, and its square has a clean interpretation as shared variance. For public health data, where variables arrive in wildly different units from wildly different sources, that comparability is worth a great deal. Answering why also protects the post from becoming a definition anyone could have copied from a textbook.

Balance that with what r assumes and what it cannot say, because a post that only praises it has not thought about it. It measures linear association only, so a genuine U-shaped relationship — the association between sleep duration and mortality is the standard public health example — can return a correlation near zero while being strong and important. It is sensitive to outliers, and in a small sample a single extreme observation can create or destroy it. It says nothing about the slope, so a strong correlation can accompany a clinically trivial change. And it is not causation, which is worth stating precisely rather than as a slogan: confounding, reverse causation and selection can each produce a strong correlation between variables with no causal link at all. Say that you would always plot the data before reporting r, because the scatterplot shows you all four problems at once.