American Community Survey of the Census Bureau; PUMS is its Public Use Microdata Sample.
ATUS
American Time Use Survey.
BIC
Bayesian information criterion, used to choose between specifications.
CVAE
Conditional variational autoencoder: a neural network that learns a distribution of households conditional on
PUMA and household size and generates new households from it.
Episode
One activity of a person on the simulated day, with a type, start time, duration and location.
GEH
A statistic comparing hourly modelled and counted flows; values below 5 are a common target.
Hold-out half
The half of the travel-survey households not used in any estimation or calibration step.
Inclusive value (logsum)
The expected maximum utility of a lower-level choice, passed up to a higher-level choice.
IPF, IPU
Iterative proportional fitting and updating: reweighting methods of population synthesis.