terms.GroupSpecificTerm

terms.GroupSpecificTerm(expr, factor)

Group-specific model term.

Group specific terms are of the form (expr | factor). The expression expr is evaluated as a model formula with only common effects and produces a model matrix following the rules for common terms. factor is inspired on factors in R, but here it is evaluated as an ordered pandas.CategoricalDtype object.

The pipe operator | works as in R package lme4. As its authors say: “One way to think about the vertical bar operator is as a special kind of interaction between the model matrix and the grouping factor. This interaction ensures that the columns of the model matrix have different effects for each level of the grouping factor”

Parameters

expr : Intercept or Term

The term for which we want to have a group specific term.

factor : Term

The factor that determines the groups in the group specific term.

Attributes

data : scipy.sparse.csr_matrix

The values associated with the term as they go into the design matrix.

kind : str

Indicates the type of the term. Can be one of "numeric", "categoric", or "interaction".

Methods

Name Description
eval_new_data Evaluates the term with new data.
eval_new_data_group_index Evaluate the grouping factor as indices for new data.

eval_new_data

terms.GroupSpecificTerm.eval_new_data(data)

Evaluates the term with new data.

Converts the variable in factor to the type remembered from the first evaluation and produces the design matrix for this grouping, calls .eval_new_data() on self.expr to obtain the design matrix for the expr side, then computes the design matrix corresponding to the group specific effect.

Parameters

data : pd.DataFrame

The data frame where variables are taken from.

Returns

Zi : scipy.sparse.csr_matrix

eval_new_data_group_index

terms.GroupSpecificTerm.eval_new_data_group_index(data)

Evaluate the grouping factor as indices for new data.

Unlike eval_new_data, this method preserves the distinction between missing values and unseen, non-missing levels. Existing factor columns are represented by their fitted index, missing values by -1, and unseen levels by consecutive indices after the fitted columns.

Parameters

data : pandas.DataFrame

The data frame where factor values are taken from.

Returns

index : numpy.ndarray

One-dimensional int64 array. Existing groups use indices in 0..G-1, missing groups use -1, and new groups use G...

new_groups : tuple

New, non-missing levels in first-occurrence order. Interaction levels are represented as tuples of their component values.