Search code examples
time-seriesstatsmodelslog-likelihood

The log-likelihood function of ARMA model after the deprecation of the statsmodels.tsa.arima_model.ARMA

I am following a tutorial building an AR model using ARMA model from statsmodels.tsa.arima_model, I got an error that it has been deprecated, and it is replaced by ARIMA from statsmodels.tsa.arima.model. I am trying to write a function that calculate a test statistic of the log-likelihood ratio test, which requires the use of the llf attribute of an initialized ARMA model. I do not know what is the counterpart of llf attribute in ARIMA model?

I tried statsmodels.tsa.arima.model.ARIMA.fit().loglike() but I did not know what kind of parameter it should get?
ar_2_model = ARIMA(data, order=(2,0,0) ).fit()

I tried ar_2_model.loglike(), but I could not figure out what kind of parameters to pass to it.

I want to get the value of the log-likelihood function of ar_2_model as similar to if I would have used
ar_3_model = ARMA(data. order=(2,0)).fit()
ll_f = ar_3_model.llf

Solution

  • You can still use llf:

    ar_2_model = ARIMA(data, order=(2,0,0) ).fit()
print(ar_2_model.llf)