ergmx.terms.Curved

class ergmx.terms.Curved(term, cutoff=30)

A geometrically weighted term whose decay is estimated, as ergm’s fixed=FALSE: a curved exponential family term.

Its statistics are the counts of the histogram the term weights: ties with exactly 1, 2, … K edgewise shared partners for gwesp, vertices with degree 1, 2, … K for gwdegree, with K the cutoff (30 by default, as in ergm, or the largest possible count if smaller). With parameters theta and decay alpha, the coefficient of count k is

eta_k = theta * exp(alpha) * (1 - (1 - exp(-alpha))^k),

so that eta . counts is theta times the term with a fixed decay alpha. If the cutoff is below the largest possible count, a last statistic counts everything above it, with coefficient theta * exp(alpha), the limit of eta_k; ergm instead stops with an error when the cutoff is exceeded.

curved = True

its statistics’ coefficients are a nonlinear function of fewer parameters (gwesp with an estimated decay).

Type:

Whether the term is curved

dyad_independent = False

Whether the term’s change statistic only depends on the toggled dyad.

property triadic

Returns True when the argument is true, False otherwise. The builtins True and False are the only two instances of the class bool. The class bool is a subclass of the class int, and cannot be subclassed.

property directed

The type of the None singleton.

property degree_dependence

The type of the None singleton.

names(network)

Names of the term’s statistics.

param_names(network)

Names of the term’s parameters: its statistics’, unless curved.

spec(network)

The term as the Rust core expects it: (name, real params, integer params).

initial()

The starting value of the decay: the decay argument.

starts(network)

Starting values of the parameters a fit can’t start at 0 (the decays of curved terms), as (position among the term’s parameters, value).

eta(params, network)

Coefficients of the statistics from the parameters (the same, unless curved).

jacobian(params, network)

Derivatives of eta with respect to the parameters (statistics x parameters).