Glossary

Adaptive biased coin design

A randomization method where the probability of assigning a participant to a treatment arm changes based on existing imbalances. The urn model by Wei (1978) is one implementation.

API key

A secret token assigned to each user that authenticates REST API requests. Generated automatically by the flask add_user command.

Imbalance measure

A statistic that quantifies how uneven the treatment allocations are within a stratum. Supported measures are range, variance, and chisquare. See Configuration for details.

PCG64

The pseudorandom number generator used by NumPy and this system. Given the same starting_seed, the sequence of assignments is fully reproducible.

Plugin

A Python module in the plugins/ directory that can override or modify treatment assignments after the urn draw. See Plugin System.

Prognostic factor

A baseline characteristic (e.g., age group, site, disease severity) used to stratify randomization so that treatment groups are balanced within each factor level.

Stratum

A subgroup defined by a specific combination of prognostic factor levels. For example, site=Site 1, sex=Male is one stratum. Each stratum maintains its own urn.

Treatment arm

One of the experimental conditions to which participants can be assigned (e.g., Treatment A, Placebo).

Urn

A conceptual container holding colored balls, one color per treatment arm. Drawing a ball determines the treatment assignment. After each draw, balls are added back according to the alpha and beta parameters, shifting future probabilities toward balance.

Urn randomization

The randomization scheme described by Wei (1978) in which an urn model adaptively adjusts assignment probabilities to reduce treatment imbalance across prognostic factors.