Week 7 - Variability & Model Fit
How Good Is This Model?
For three weeks we have been reading estimates — slopes, intercepts, group differences. This week we read the other half of the output: the numbers that tell you whether the model is worth anything at all.
The core idea is simple arithmetic. Start with how wrong you would be using nothing but the mean. Fit a line. Whatever distance the line covers is the model; whatever is left is the residual. That single split gives you R², the F test, and the residual standard error — three numbers you have been glancing past since Week 5.
“Statistically significant” and “explains a lot” are different claims. With 400 people you can be very confident about a slope that accounts for 6% of the variance. Both facts are true, and a careful results section says both.
We will also meet the first real trap in model building: R² can never go down when you add a predictor, even a pointless one. That is why adjusted R² exists, and why it is the number to watch once your models have more than one thing in them.
In-Class Check 2 runs at the start of class — about 20 minutes, open notes, open readings, open past labs. It covers Weeks 5 and 6: slopes, intercepts, prediction and residuals, R², and groups as a line.
Prepare
Slides
💻 Variability & Model Fit - JAMOVI
Both decks use the same data and reach the same numbers. JAMOVI has one genuine advantage this week - the Model Builder compares nested models and reports ΔR² for you - so the JAMOVI deck spends a little longer there.
Data
Lecture: ASAP Sleep & Mood Data (.csv)
In-class activity: Spotify Data (.csv)
In-Class Activity
For Next Time
📋Lab 8 - How Good Is This Model? — due Tuesday 10/13 at 11:59pm
Monday 10/12 and Tuesday 10/13 are Fall Break — no classes. Lab 8 is still due Tuesday at 11:59pm and is submitted online as usual. There are no office hours that week before the deadline, so come by Tuesday 10/6 or email me if you want help before the break.
📖Chapter 10 - ROS — multiple regression, for next week
The Midterm Project is posted now and is due Tuesday 10/27.
Parts 1-5 use only what you already know, so you can start this weekend. Part 6 needs multiple regression, which we cover on 10/15. No lab is due in Weeks 9 or 10, so that time is yours for the project.
Back to course schedule ⏎