AR1 in proc autoreg vs. Lag(y) in proc reg

Taliah — Wed, 19 Feb 2020 15:00:51 GMT

We ran Proc autoreg. AR1 (autoregressive term 1) was eliminated. It's p value = 0.06 (all other autoregressive terms we tried were also not significant).

We then ran proc reg on the same data set. Darbin-Watson indicated we have autocorrelation in the data.

We added lag(y) as a dependent variable in proc reg (i.e., lag of the dependent variable), and it was highly significant (p=0.0004)

In my understanding if AR1 in proc autoregressive is clearly not significant, then lag(y) in proc reg ahuols also be not significant. Would appriciate any insight on this, and what we can check / what shuold be done in such a case.

There are 4 other dependent variables in the model, all numeric. Our question is regarding lag(y) vs. AR1

Thank you,

Tali

topic AR1 in proc autoreg vs. Lag(y) in proc reg in Statistical Procedures

AR1 in proc autoreg vs. Lag(y) in proc reg