Exercise 2: Why the Direction of a Correlation Differs While the Mechanism Doesn't — Possible Solution ==================================================================== WHAT ds1-6's PEARSON MECHANISM ACTUALLY MEASURES ------------------------------ Per ds1-6, Pearson's correlation coefficient measures "how strongly two variables move together," ranging from −1 (perfect inverse movement) to 1 (perfect matching movement). The mechanism itself is entirely neutral about which DIRECTION the relationship happens to run — it simply measures whatever direction the two variables actually move in together, positive or negative, based on the real data provided to it. WHY quantity-vs-revenue (ds1-7) MOVES POSITIVELY ------------------------------ In the coffee-shop dataset, ordering a larger quantity of a product directly and mechanically produces higher revenue — more items sold at a roughly consistent per-item price means the total naturally increases alongside quantity. There's a direct, intuitive real-world reason these two specific variables would rise and fall together, producing a positive correlation. WHY mileage-vs-price (this chapter) MOVES NEGATIVELY ------------------------------ In the used-car dataset, higher mileage reflects a more heavily used vehicle, which real-world car buyers and sellers consistently value LESS than a lightly used one — so as mileage goes up, price tends to go down. This is a completely different real-world relationship between a completely different pair of variables, with its own independent reason (wear and reduced remaining lifespan) for moving in the opposite direction from the coffee-shop example. WHY THIS DOESN'T MEAN THE MECHANISM ITSELF DIFFERS ------------------------------ Per ds1-6's own definition, Pearson's coefficient doesn't presuppose any particular direction — it's built to detect and report whichever direction the real relationship actually has, positive, negative, or none at all. The fact that one dataset happens to show a positive relationship and another happens to show a negative one is a fact about the real-world phenomena being measured (how quantity relates to revenue vs. how mileage relates to price), not a fact about the statistical tool itself behaving differently. The same formula, applied faithfully to two different real relationships, correctly reports two different signs — that's the tool working exactly as designed in both cases, not evidence of two different underlying mechanisms. WHY THIS WORKS AS AN ANSWER ------------------------------ It explains the specific real-world reason each dataset's own two variables move in the direction they do, and separately explains why Pearson's own correlation mechanism (per ds1-6) is direction-neutral by design — meaning two opposite-signed results from two different datasets are both correct applications of the identical underlying tool, not evidence that the tool itself changed between chapters.