OMBA 102 · Unit 5 of 7
Relationships and Prediction
Data, Statistics and Managerial Decisions
Start unit · 5 lessons →Learning objectives
After completing this unit, you will be able to:
- Interpret correlation and its limits for managerial decisions
- Build and read simple and multiple regression models
- Validate forecasts and measure prediction error honestly
- Distinguish correlation from causation and design better tests
- Apply experimental thinking when observational data misleads
Why this matters
"Drivers of growth" regressions and forecast decks influence budgets and strategy. Unit 5 teaches you to use relationships and prediction tools without confusing fit with causation or overfitting history.
Unit overview
Work through the five lessons below in order. Replicate regression examples in a spreadsheet.
| # | Lesson | Core idea |
|---|---|---|
| 1 | Correlation and Its Limits | Association vs mechanism |
| 2 | Simple Linear Regression | Line fit, slope, intercept |
| 3 | Multiple Regression | Controls and interpretation |
| 4 | Forecast Accuracy and Model Validation | Train/test, error metrics |
| 5 | Causality, Confounding, and Experimental Design | When prediction is not proof |
Connection to applied work
Add a regression or forecast section to your project with explicit limits: what you predict, what you cannot claim causally, and validation approach.
Practice
- Interpret a correlation table without causal language.
- Fit a simple regression; explain slope in business units.
- Add one control variable in multiple regression and note coefficient change.
- Compare in-sample vs out-of-sample forecast error.
Knowledge check
- Why does correlation not imply causation?
- What does R-squared measure and overstate?
- What is confounding?
- How do you know a forecast is overfit?
- When is an experiment required?
Key takeaways
- Relationships help prediction and hypothesis generation, not automatic policy.
- Validation discipline separates models from stories.
- Causal claims need design, not just more variables.
- Complete lessons before assessments.
Unit assessment
Complete each section below. Score 80%+ on the quiz to finish this unit's assessment.
Exercises
Apply what you learned in this unit with structured practice.
Deliverable
300–500 word analysis document saved to your portfolio under OMBA 102.
Rubric
- • Framework applied correctly (not just named)
- • Specific evidence from a real example
- • Clear recommendation with tradeoffs acknowledged
- • Professional writing with source citation
Deliverable
Problem solutions + 150-word reflection in your OMBA 102 workbook.
Rubric
- • Attempted all practice items before checking answers
- • Honest reflection on errors
- • Identifies a specific review action
Model / spreadsheet
Build or extend a spreadsheet model tied to this unit.
Deliverable
Structured model document (2+ pages) · One-paragraph summary of key insight from the model · Screenshot or export saved to portfolio
Rubric
- • Assumptions stated explicitly
- • Logic is auditable (formulas or steps visible)
- • Output answers a specific business question
- • Sensitivity or scenario considered
Knowledge quiz
Check your understanding before marking the unit complete.
1. Support contacts correlate r = 0.72 with retention, but contacts respond to distress. The best managerial read is:
2. If r = 0.60 between ad spend and revenue, what is r²?
3. In simple linear regression of revenue on ad spend, a slope of 4.2 means:
4. Adding customer tenure to a churn model changes the sign on campaign exposure. This most likely indicates:
5. A model with 40 predictors fit to 45 historical weeks will likely:
6. MAPE of 8% on a demand forecast means, roughly:
7. To estimate causal effect of a new pricing page, the strongest design is:
8. Extrapolating a linear trend in ad spend to double current levels primarily risks: