Source code for allensdk.internal.model.AIC
import numpy as np
"""
TODO: license
TODO: comment style
"""
[docs]def AIC(RSS, k, n):
"""
Computes the Akaike Information Criterion.
RSS-residual sum of squares of the fitting errors.
k - number of fitted parameters.
n - number of observations.
"""
AIC = 2 * k + n * np.log( RSS/n)
return AIC
[docs]def AICc(RSS, k, n):
"""
Corrected AIC. formula from Wikipedia.
"""
retval = AIC(RSS, k, n)
if n-k-1 != 0:
retval += 2.0 *k* (k+1)/ (n-k-1)
return retval
[docs]def BIC(RSS, k, n):
"""
Bayesian information criterion or Schwartz information criterion.
Formula from wikipedia.
"""
return n * np.log(RSS/n) + k * np.log(n)