I am following the course, and in the video 5:07, the instructure says that the ARMA(5,6) has significant coefficients, but i test it in my side and it didn’t work.:
model_ar_5_ma_6 = ARMA(df.market_value, order = (5,6))
results_ar_5_ma_6 = model_ar_5_ma_6.fit(start_ar_lags = 7)
Do you know why?
When we first wrote the code (and the text accompanying it), statsmodels was on version 0.9.
With some of the updates came changes to the default values for some method, which lead to different results from what we had initially. Essentially, we’re slightly worried to re-record the lectures where this appears to be the case, since another update can once again switch the default outputs and bring us back to square one.
We’re thinking of the best ways to deal with this issue. The easy fix is to ask all students to use an outdated version of the package just to make our videos seem flawless. However, it’s best to always use the latest version of every package (whenever possible), so we’re currently in limbo until we decide what course of action to take.