macroeco.compare.AIC

macroeco.compare.AIC(data, model, params=None, corrected=True)

Akaike Information Criteria given data and a model

Parameters:

data : iterable

Data for analysis

model : obj

Scipy frozen distribution object. When freezing, keyword args loc and scale should only be included if they represent a parameter.

params : int

Number of parameters in the model. If None, calculated from model object.

corrected : bool

If True, calculates the small-sample size correct AICC. Default True.

Returns:

float :

AIC(C) value

Notes

AICC should be used when the number of observations is < 40.

References

[1]	Burnham, K and Anderson, D. (2002) Model Selection and Multimodel Inference: A Practical and Information-Theoretic Approach (p. 66). New York City, USA: Springer.

Examples

>>> import macroeco.models as md
>>> import macroeco.compare as comp

>>> # Generate random data
>>> rand_samp = md.nbinom_ztrunc.rvs(20, 0.5, size=100)

>>> # Fit Zero-truncated NBD (Full model)
>>> mle_nbd = md.nbinom_ztrunc.fit_mle(rand_samp)

>>> # Fit a logseries (limiting case of Zero-truncated NBD, reduced model)
>>> mle_logser = md.logser.fit_mle(rand_samp)

>>> # Get AIC for ztrunc_nbinom
>>> comp.AIC(rand_samp, md.nbinom_ztrunc(*mle_nbd))
765.51518598676421

>>> # Get AIC for logser
>>> comp.AIC(rand_samp, md.logser(*mle_logser))
777.05165086534805

>>> # Support for for zero-truncated NBD over logseries because AIC is
>>> # smaller

>>> # Call AIC with params given as 2 (should be the same as above)
>>> comp.AIC(rand_samp, md.nbinom_ztrunc(*mle_nbd), params=2)
765.51518598676421

>>> # Call AIC without sample size correction
>>> comp.AIC(rand_samp, md.nbinom_ztrunc(*mle_nbd), params=2, corrected=False)
765.39147464655798