I was just trying to fit a model on stock prices. First i tried Ar(1) model then MA(1) model then Arima(1,1) model .
However the Maximum Log Likelihood of all the three models was very similar to each other.
Does it mean that using extra variables(parameters) doesnt add to the value of the model and i am better off using the AR(1) model ? Or something else has to be interpreted that i am missing.
Is there a accepted threshold to judge the maximum log likelihoods of two models?