Automating lag length selection **not unit root testing**
Posted: Wed May 01, 2013 1:11 am
Hi there,
I'm trying to run some relatively simple regressions of the form: y c x(0 to -a), but over about 20 different versions of y and x.
I would like to automate the procedure of selecting length of lags, rather than choosing a generic number for all (or specifying 20 different lag lengths). I've searched this forum (and elsewhere) for automatic lag length selection, but everything seems to relate to cointegration or unit root testing, while I'm trying to do something more simple. But surely, if it's possible to automate lag length selection when testing for a unit root (presumably by minimising the selected information criterion), it's possible to do the same for multivariate equations?
Any thoughts would be appreciated. Thanks!
I'm trying to run some relatively simple regressions of the form: y c x(0 to -a), but over about 20 different versions of y and x.
I would like to automate the procedure of selecting length of lags, rather than choosing a generic number for all (or specifying 20 different lag lengths). I've searched this forum (and elsewhere) for automatic lag length selection, but everything seems to relate to cointegration or unit root testing, while I'm trying to do something more simple. But surely, if it's possible to automate lag length selection when testing for a unit root (presumably by minimising the selected information criterion), it's possible to do the same for multivariate equations?
Any thoughts would be appreciated. Thanks!