An approach to model selection in which one bases inference on an average of
all possible models instead of a single best model. The BMA
is largely used in various branches of knowledge
to properly account for model uncertainty in performing predictions.
See also: → Bayesian; → model;
→ average.