Data dictionary

This appendix documents all datasets in the data/ folder. These files are pre-compiled and used throughout the book to avoid requiring model refits or long-running simulations during rendering.

sim_dat.csv

The primary simulated dataset used in Planning and Data processing. It is synthetic, generated to match the structure of previously collected data in the lab. It is not real data.

Variable	Type	Values	Description
pid	character	P001–P216	Participant identifier
item_id	integer	1–32	Item identifier
group	factor	control, embodied	Experimental condition (between-person)
test_time	factor	pre, post, delay	Testing occasion
correct	integer	0, 1	Binary accuracy on each trial
pre_score_z	double	continuous	Standardised pre-test score

sim_dat_wide.csv

A wide-format version of sim_dat.csv, used in Data processing as a toy example for demonstrating pivot_longer(). Each row is one participant × test time combination, with one column per item.

Variable	Type	Values	Description
pid	character	P001–P216	Participant identifier
group	factor	control, embodied	Experimental condition (between-person)
test_time	factor	pre, post, delay	Testing occasion
pre_score_z	double	continuous	Standardised pre-test score
item_1 … item_32	integer	0, 1	Binary accuracy, one column per item

dat.csv

A simplified version of sim_dat.csv used in Modelling and Inference. It retains only the post-test observations and drops variables not needed for the modelling chapters, giving a cleaner starting point for readers focused on model fitting rather than data processing.

Variable	Type	Values	Description
pid	character	P001–P216	Participant identifier
item_id	integer	1–32	Item identifier
group	factor	control, embodied	Experimental condition (between-person)
correct	integer	0, 1	Binary accuracy on each trial
pre_score_z	double	continuous	Standardised pre-test score

p_100_sims.csv

The output of the simulation study in Planning (section 6.6). Each row is one parameter estimate from one fitted model across the simulation grid. This file is read in by the plotting and precision chunks, which render without needing to rerun the simulations.

Variable	Type	Values	Description
rep	integer	1–100	Simulation replicate index
subj_n	integer	100, 150, 200	Number of subjects in that simulation
b1	double	0.15, 0.40, 0.87, 1.39	True group effect (log-odds) used in data-generating process
parameter	character	Intercept, groupembodied, pre_score_z	Model parameter name
Estimate	double	continuous	Posterior median for that parameter
Est.Error	double	continuous	Posterior SD
Q2.5	double	continuous	2.5th percentile of posterior (lower 95% CI)
Q97.5	double	continuous	97.5th percentile of posterior (upper 95% CI)

Models

Pre-fitted model objects are stored in models/ as .rds files. All were fit with brms using CmdStan on the backend.

File	Description	Chapter
fit_b1.rds	Bernoulli multilevel model fit to sim_dat.csv (post-test only)	Planning
sim_b1.rds	Same formula as fit_b1, fit to a single simulated dataset (d1) to validate the data-generating process	Planning
fit_b2.rds	Bernoulli multilevel model fit to dat.csv, the wrangled dataset used in the modelling and effects chapters	Modelling