I’m interested in generating a PACF-like plot but between two separate time series x and y, with significance levels included.
Just as the PACF tells us the best lagged predictors of y using lagged version of y, I’d like to know the best lagged predictors of y given lags of x without having to do this manually.
My end goal is to say, for example, “y at time t is best predicted by x in period t-1 and t-5.”
I’ve also thought of just running the cross correlations to reach this conclusion, but given my reading it seems best to find the lags with highest direct effect (PACF) vs. total effect (ACF). I appreciate any suggestions and corrections to my thinking.