gfa.analysis

Compute descriptive statistics related to group factor analysis (GFA) models.

Functions

predictive_performance

Compute the leave-group-out predictive performance of a GFA model.

---

predictive_performance(

obs_data: ObsStatic,

obs_posterior: ObsParamsPosterior,

y_dims: ndarray | None = None,

) → tuple[float, float]

Compute the leave-group-out predictive performance of a GFA model.

Parameters:

obs_dataObsStatic

Observed data.

obs_posteriorObsParamsPosterior

Fitted observation model posterior.

y_dimsndarray or None, default None

Dimensionalities of each observed group. If None, inferred from obs_data.

Returns:

R2float

Leave-group-out R^2.

MSEfloat

Leave-group-out mean squared error.

Examples

>>> from latents.gfa import GFAModel
>>> from latents.gfa.analysis import predictive_performance
>>> model = GFAModel()
>>> model.fit(Y)
>>> R2, MSE = predictive_performance(Y, model.obs_posterior)