Time series regressions

[msd86]

Hello everyone.

I have two main time series A and B and have found them to be cointegrated and cross correlating at lag 4.

I run a regression model by inserting variables A (lagged at 4) along with other 3 time series to explain variable B. I made sure all variables are pre-whitened.

But none of the variables was significant (which was strange in view of their logical relation). how can i explain that A and B are both cross-correlating at lag 4 but that their regressions coefficients are not significant? am trying to answer to the question of whether they are significantly related.

 

[noetsi]

What type of regression are you running? There are many types of time series regression and each has their own rules.

Supposedly vector error correction models are run with cointegration. Or that is the state of the art. Sound like you are doing multivariate arima instead since you are pre-whitening.