ARMA model

Each group has been assigned a time series, stored in the file groupXX_arima.dta. It contains 1,000 observations of the variable yt.
Test whether there is a unit root in the time series.
Estimate the AR(4), ARMA(3,1) and ARMA(2,2) models. Report the results in a table with one column for each model. For each model, include the Akaike and Schwarz Bayesian Information Criteria, as well as the Ljung-Box test of white noise of the first four autocorrelations.
Compare the performance of the different models. Using both the Akaike and Schwarz Bayesian Information Criteria, which of the three models describes the variable yt best? What is the conclusion of the Ljung-Box test?
If you are not happy with the performance of these four aforementioned models, consider whether to estimate an alternative model of your choice.