I ran an ARDL on my data that yielded short run coefficients that are significant in say lag(1), not lag(2) and again in lag(3) While for other variables it is significant in lag(3) only. How am I supposed to interpret.
An economist - like any scientist - should have control over the method, i.e. should not be controlled by the method. Before applying a method, it is important to have a theoretical model. Then it is easier to interpret the results of regressions. Coefficients which are totally against any theory of logic should not be accepted, even if they are significant.