Exercise 2: Why Department's Own Apparent Effect Is a Genuine Confounding-Variable Question — Possible Solution ==================================================================== WHAT THE HYPOTHESIS ACTUALLY ASKS ------------------------------ Per this chapter, Step 7 raises: "Does department matter, or is department's own apparent effect actually explained by department-level salary differences instead?" — explicitly named as "a genuine, unresolved confounding-variable question, straight out of ds1-6's own material." WHAT ds1-6's CONFOUNDING-VARIABLE MATERIAL ESTABLISHES ------------------------------ Per ds1-6, "ice cream sales and drowning incidents correlate strongly... Neither causes the other; a third, unmeasured factor... is the real driver." The general lesson: two things can appear related (or one can appear to explain an outcome) purely because both are actually driven by a third factor, rather than one directly causing the other. WHY DEPARTMENT COULD BE EXACTLY THIS KIND OF CASE ------------------------------ Looking at this chapter's own sample table, Sales employees (E01, E03, E06) have noticeably lower salaries than Engineering employees (E02, E04, E07), and Sales also has a higher share of employees who left (E01 and E06, both Yes). If department appears associated with attrition, it's entirely possible that department itself has no direct causal effect on someone's decision to leave at all — what actually drives attrition might be salary alone, and department only appears to matter because departments happen to pay differently. In that case, department would be a stand-in for salary, not an independent cause in its own right — precisely the confounding-variable pattern ds1-6 described with ice cream and drowning both really being driven by summer heat. WHY THIS CAN'T SIMPLY BE TRUSTED AT FACE VALUE ------------------------------ Nothing in this chapter's own Steps 1-6 actually separates "department's own independent effect" from "salary's effect, which happens to correlate with department." The box plot in Step 4 showed salary differing by whether someone left; if salary also differs systematically by department, then department's own apparent association with attrition could be entirely explained away by controlling for salary — or it might not be, and department could still matter independently. Nothing performed in this capstone actually tests which explanation is correct; the pattern is only VISIBLE, not yet disentangled. WHY THIS IS "GENUINE AND UNRESOLVED," NOT JUST A LABEL ------------------------------ The chapter deliberately doesn't attempt to resolve this question itself — resolving it would require exactly the kind of controlled, multi-variable modeling technique this course's own scope note reserves for ml1. Naming it as unresolved here is honest about the real limit of what EDA alone can determine, rather than overclaiming a conclusion the data collected so far can't actually support. WHY THIS WORKS AS AN ANSWER ------------------------------ It applies ds1-6's own confounding-variable pattern directly to the department/salary relationship visible in this chapter's own sample data, explains concretely why department could be a stand-in for salary rather than an independent cause, and explains why nothing performed in this capstone actually settles which explanation is correct.