Quantitude S7E13 Time-Varying Effect Models
Jan 20, 2026
Dive into the fascinating world of time-varying effects models with unique statistical insights. Discover how relationships between predictors and outcomes shift over time, illustrated with poignant examples like social support and mental health. The hosts break down complex ideas into engaging visuals, exploring the balance between rigid and flexible modeling. They also tackle limitations and practical applications. Prepare for laughter as they throw in whimsical references to pop culture and everyday quirks, making statistics entertaining!
AI Snips
Chapters
Books
Transcript
Episode notes
Cal Ripken To Zen: A Driving Example
- Patrick Curran shares becoming less reactive while driving as he aged, illustrating changing relations over time.
- He uses this personal change to motivate thinking about how predictors' effects can change with age.
Plot Relations As Functions Of Time
- Time-varying effect models (TVEM) plot a regression coefficient across time to show how the relation between a predictor and outcome changes.
- TVEM lets the coefficient vary smoothly over age or time instead of forcing a single fixed or linear effect.
Limits Of Parametric Time Interactions
- Traditional multilevel models estimate a time-varying covariate as constant unless explicitly interacted with time.
- Interacting with time yields a linear or polynomial change, which may miss complex nonparametric patterns.
