Hi there,

When using a first difference regression a key assumption is that there is no serial correlation in the errors. By construction there is an MA(1) process error so I need to test for evidence of second-order serial correlation. When performing this test should I specify the regular model without first difference i.e.
Code:
xtserial y x1 x2 x3, output
or use it in it's first difference form?

Code:
xtserial d.y d.x1 d.x2 d.x3, output
Thank you for any help