Pacfm acf

Why do we take Pacf for AR and Acf for Ma to find the optimal lag. why pacf is preferred for auto regressive models and why acf is considered for moving average model, why not vice versa

Hi Jagteshwar,



PACF is suitable for AR models because it shows the direct relationship between the series and its lags, cutting out the noise from intermediate lags. This makes it easier to pinpoint the exact number of lags (p) needed in an AR model. On the other hand, ACF works better for MA models since it captures the overall correlation pattern, which reflects the lagged error terms. If you swapped them, you’d get a less clear picture - ACF can’t isolate direct effects well for AR, and PACF doesn’t capture the error-driven structure of MA models. That’s why we use PACF for AR and ACF for MA. 



Hope this helps!