Start with a simple model and expand it !
- Significance coefficient
- Simpler model is of more practical use than a complex model unless the complex model gives significantly better results.
- Log Likelihood ratio test : only applied with different degree of freedom.
AIC : Akaika information coefficient
BIC : Bayesian Information coefficient
opt for the model with less AIC and BIC coefficents
- Residuals follow normal distribution with mean 0
- If comparing two models go for the model with higher Log likelihood and lower IC.