

AR effects with lagged() vs. without ... 

Message/Author 


I'm curious about the analytic differences between specifying an AR model with vs. without using the lagged command. Let's say we wanted to run an AR2 model. It would seem that one can specify it like so, using the "lagged" command: VARIABLE: usevariables = y; lagged = y(2); MODEL: %within% AR1  y ON y&1; AR2  y ON y&2; However, it would seem that one could specify the same model if the two lagged variables (y_1 and y_2) already existed in one's data set, *without* using the "lagged" command: VARIABLE: usevariables = y y_1 y_2; within = y_1 y_2; MODEL: %within% AR1  y ON y_1; AR2  y ON y_2; Are there any "under the hood" analytic differences in these two approaches? Thanks! 


Yes. Y&1 = Y_1  YB where YB is the random intercept/between part of Y. So Y&1 is centered while Y_1 is not. Another difference is how the first two periods are treated. 


Thanks! 

Back to top 

