Are autocorrelations for time series model the correlation between the error term for time t and t-1, t-2, t-3, and so on or are they correlations between t and t-1, t-1, and t-2, t-2 and t-3. Please Let me know thanks!
The correlation between t & t-1 is one and the same as t-1 and t-2. So that is ruled out.
It has to be between t and t-1 or t and t-2 or t and t-3 etc …
You need to understand the concept. What we are doing is finding the correlation between an error term at t and the error term at t-1 , when we are looking at an autocorrelation with a lag of one.
So by the same logic, an autocorrelation for a lag of 2 means , you have a correlation of error term at t and error term at t-2.
Hope this helps 