Bayesian Multilevel Models for Repeated Measures dаta A Conceptual and Practical Introduction in R

This book offers a hands-on, conceptual introduction to Bayesian multilevel models for analyzing repeated measures data, a common data type in linguistics, psychology, and cognitive science. Starting with simple models and progressing to more complex ones like multinomial regression, the authors use a single, realistic experimental dataset throughout to provide fully worked examples in R using the `brms` package. Instead of getting bogged down in mathematical theory, the book focuses on building intuitive, geometric understanding and practical coding skills, making it accessible for readers with any level of statistical background who want to move beyond traditional methods and harness the flexibility of Bayesian modeling for their own research.

What it argues

Bayesian Multilevel Models for Repeated Measures dаta A Conceptual and Practical Introduction in R

Key ideas it contributes

• Fundamental Frequency (f0) — The rate of vibration of a speaker's vocal folds, measured in Hertz (Hz). It is the primary acoustic correlate of perceived voice pitch and serves as a cue for speaker size, age, and gender.
• Acoustic Vocal-Tract Length (VTL) — The effective acoustic length of a speaker's vocal tract, estimated in centimeters from the resonant frequencies (formants) of the speech signal. It is a primary acoustic correlate of perceived speaker size and body length.
• Apparent Age — The listener's categorical judgment of whether a speaker is a child or an adult. This perceptual categorization acts as a moderator, changing how acoustic cues are interpreted to judge height.
• Apparent Gender — The listener's categorical judgment of whether a speaker is male or female. This perceptual categorization also acts as a moderator for how acoustic cues are used.
• Apparent Height — The listener's estimate of a speaker's physical height, in centimeters. This is the primary outcome variable, influenced by acoustic cues and the listener's perceptual categorizations.
• Listener Variation — Systematic, idiosyncratic differences between individual listeners in their average judgments (intercepts) and their use of acoustic and categorical cues (slopes). This is a key source of non-independence in repeated measures data.
• Speaker Variation — Systematic, idiosyncratic differences between individual speakers' voices that influence listener perceptions but are not captured by f0 and VTL alone. This is another source of non-independence.