Capstone: A Full EDA Project on a Real Dataset

8 rows, 6 columns. One categorical (department), one boolean-like categorical (left_company), four numeric. No missing values in this small sample — unusual for a real dataset, and worth naming rather than assuming.

Salary ranges roughly $44,000–$96,000, mean around $65,000. Years at company ranges from 0.5 to 14 — a wide spread worth investigating directly in Step 3.

df["years_at_company"].plot(kind="hist")
df["department"].value_counts()
df["left_company"].value_counts()

✓ Finding: years_at_company skews heavily toward short tenures — most employees in this sample have been at the company under three years.

df.plot(kind="scatter", x="years_at_company", y="salary")
sns.boxplot(data=df, x="left_company", y="salary")

✓ Finding: a box plot of salary grouped by left_company shows the "Yes" group's salaries sitting visibly lower than the "No" group's — a real, visible bivariate pattern, not yet a conclusion.

sns.heatmap(df[["age","years_at_company","salary"]].corr(), annot=True)
sns.pairplot(df[["age","years_at_company","salary"]])

✓ Finding: age and years_at_company correlate strongly and positively with each other — unsurprising, since both track roughly the same underlying "how established" idea — while salary correlates more weakly with either alone.

Nothing in this small sample clears the 1.5×IQR threshold — a genuinely different, and equally valid, outcome from ds1-9's own used-car dataset, where the Jaguar was flagged immediately. A real EDA process sometimes finds nothing anomalous, and that absence is itself a legitimate, worth-recording finding, not a failed step.

• "Does lower salary predict a higher likelihood of an employee leaving, once tenure is accounted for?" — directly motivated by Step 4's own box-plot finding.
• "Is short tenure itself a risk factor for leaving, independent of salary?" — directly motivated by Step 3's own skewed-tenure finding.
• "Does department matter, or is department's own apparent effect actually explained by department-level salary differences instead?" — a genuine, unresolved confounding-variable question, straight out of ds1-6's own material.