# Information Criteria

Information Criteria are used in Q for comparing models in situations where statistial tests are not generally considered appropriate (e.g., stepwise regression and comparison of latent class models and trees).

The Information Criteria are ordered from the one which favors the most segments being created (AIC) to the one that favors the least being created (BIC). Where User specified is selected, the Information Criterion is -2*LL+k*P, where LL is the log-likelihood, k is the value specified (which by default is 2, which is the AIC) and P is the number of parameters.

The formulas used in Q are:

• AIC: 2*P - 2*LL
• AIC3: 3*P - 2*LL
• BIC: ln(n)*P - 2*LL
• CAIC: ln(n+1)*P - 2*LL