When doing a time series analysis I have read these instructions :
Step 1 — Check stationarity: If a time series has a trend or seasonality component, it must be made stationary before we can use ARIMA to forecast. . Step 2 — Difference: If the time series is not stationary, it needs to be stationarized through differencing. Take the first difference, then check for stationarity. Take as many differences as it takes. Make sure you check seasonal differencing as well. Step 3 — Filter out a validation sample: This will be used to validate how accurate our model is. Use train test validation split to achieve this Step 4 — Select AR and MA terms: Use the ACF and PACF to decide whether to include an AR term(s), MA term(s), or both. Step 5 — Build the model: Build the model and set the number of periods to forecast to N (depends on your needs). Source - https://towardsdatascience.com/time-series-forecasting-arima-models-7f221e9eee06
My questions are on steps one and two.
If I remove the trend and seasonal pattern, what is left of my data? According to this source, after subtracting a trend (calculated with a MA on the frequency of pattern per year), and subtracting a seasonal pattern, what is left is a random component. https://anomaly.io/seasonal-trend-decomposition-in-r/
If this is true, does this mean ARIMA is modeling the random component?
Also, what is the difference between removing the trend and differencing the data, do they have the same effect?
Furthermore, do I need to remove the seasonal component before looking at the ACF, PACF to determine the MA and AR components?
Finally, say I know there is a seasonal pattern. Does that mean I should do a SARIMA(p,d,q,P,D,Q) with D (the seasonal difference) equal to the frequency of the pattern per year? Or should I subtract the trend and seasonal pattern, and do an ARIMA. Would data ever not be stationary if one subtracts the trend and seasonal pattern?