AIC, BIC, Bayesian evidence and a notion on simplicity of cosmological model
Marek Szydłowski, Aleksandra Kurek · 2008
Recent astronomical observations indicate that the Universe is in the phase of accelerated expansion. There are many cosmological models which explain this phenomenon, but should we prefer those models over the simplest one – ΛCDM model? According to the Occam’s razor principle if all models describe the observations equally well we should prefer the simplest one. We consider the model comparison methods which involve such rules: the Akaike information criterion (AIC), Bayesian information criterion (BIC) and Bayesian evidence to compare the ΛCDM model with its generalisation where the interaction between dark matter and dark energy is allowed. The analyses based on the AIC and Bayesian evidence indicate that there is only a weak evidence in favour of the ΛCDM model over its generalisation, while those based on BIC quantity indicate the strong evidence in favour the simpler model. We also calculate some quantity which measure the effective number of model parameters that the given data can constrain. This value is used to compare the concordance LCDM model with its generalization basing on the extended interpretation of continuity condition-interacting ΛCDM cosmology. We conclude that data set are not enough informative to constrain all parameters allows to vary in the models. ∗Electronic address: [email protected] †Electronic address: [email protected]