InformationCriteria#

class ns_sdn.summary.InformationCriteria(aic, bic, hqic)#

Model-selection criteria computed from a fitted log-likelihood.

For a log-likelihood \(\ell\), \(p\) parameters and an effective sample of \(m\) observations,

\[\mathrm{AIC} = -2\ell + 2p, \qquad \mathrm{BIC} = -2\ell + p\log m, \qquad \mathrm{HQIC} = -2\ell + 2p\log\log m.\]

Lower is better for each. For a network, \(p\) is the nominal weight count and \(\ell\) a concentrated likelihood at a penalized optimum, so the criteria order fits that share an engine configuration and differ in \(K\); they do not compare a network against an ARIMA.

Variables:

  • aic – Akaike information criterion.

  • bic – Bayesian (Schwarz) information criterion.

  • hqic – Hannan-Quinn information criterion.

Attributes#

aic

Alias for field number 0

bic

Alias for field number 1

hqic

Alias for field number 2

Methods#

count(value, /)

Return number of occurrences of value.

from_likelihood(llf, nobs, n_params)

Compute all three criteria.

index(value[, start, stop])

Return first index of value.