I am trying to make autoregressive model but AutoReg documentation confusing me.
It says
Period: The period of the data. Only used if seasonal is True. This parameter can be omitted if using a pandas object for endog that contains a recognized frequency.
Maybe because of my language barrier(or statistical barrier), I can't get it clearly.
Let's say I have 12 data. Every previous 3 data are independent variable, and 4th is dependent. ex) 3,8,12,500, 4,7,11,480, 5,13,19,600.
I thought the period should be 4, and lags should be 3, since t4 consists of t1,t2,t3 and t8 consist of t5,t6,t7 and so on. However, it seems like lags 3 indicates that t13 consists of t1,t5,t9 and t14 consists of t2,t6,t10.(What chatGPT says..)
What should I do?
Besides, I can't use linear regression since features are too autocorrelated, so rmse of model is unstable.
Thanks in advance!
since period and lags are increasing, so does aic,hqic, and bic. Isn't it strange? more feature makes worse.