Least Squares & Autocorrelation
Posted: Tue Oct 12, 2010 1:27 pm
I am having great success developing an econometric model for interest elasticity and excess reserves, but I need some advice on how to estimate my standard errors.
My data is time-series. Because of this I have been estimating least squares with Newey-West -- ls(n) -- standard errors. I would like to test for auto correlation. Do I stop estimating Newey-West -- i.e. drop the n in ls(n) -- or do I keep doing Newey-West and also consider an autocorrelation scheme by adding AR(1)? I do not know enough about least sqaures standard error estimation to know the answer to this question.
Also, how should the AR(1) coefficient be interpreted? If it is insignificant can I drop the autocorrelation assumption from my model? If it significant how should the coefficient (which is known as the rho statistic if I am not mistaken) be interpreted?
Thank you for your help.
I am using E-Views 7.
- Eric
My data is time-series. Because of this I have been estimating least squares with Newey-West -- ls(n) -- standard errors. I would like to test for auto correlation. Do I stop estimating Newey-West -- i.e. drop the n in ls(n) -- or do I keep doing Newey-West and also consider an autocorrelation scheme by adding AR(1)? I do not know enough about least sqaures standard error estimation to know the answer to this question.
Also, how should the AR(1) coefficient be interpreted? If it is insignificant can I drop the autocorrelation assumption from my model? If it significant how should the coefficient (which is known as the rho statistic if I am not mistaken) be interpreted?
Thank you for your help.
I am using E-Views 7.
- Eric