Selecting Lag Orders for Autoregressive Distributed Lag Models

Summary

The document asks how to select the autoregressive order and distributed-lag order in an ADL(p,q) model, preferably through an automated workflow in a statistical package. The author is looking for a working example that uses information criteria such as AIC and BIC to choose lag lengths and then constructs the corresponding regression without manually adding and removing regressors.

No algorithm, code, data, or empirical findings are supplied; the text is a request for implementation guidance. The intended application is to examine distributed lags across several assets. A complete workflow would need to specify the candidate lag ranges, model constraints, and how criteria are compared, but those choices are not resolved here. The document is useful for identifying an econometric model-selection task, not as a reproducible procedure or recommendation.

Key ideas

• An ADL model combines lags of the dependent variable with distributed lags of explanatory variables.
• The author seeks an automated approach to selecting both lag orders.
• AIC and BIC are proposed as possible model-selection criteria.
• The document requests a working example and supplies no code or selection results.

Tags

# Autoregressive distributed lag models ADL(p,q) howto in preferably matlab (stata/R/python/C# etc)


# Autoregressive distributed lag models ADL(p,q) howto in preferably matlab (stata/R/python/C# etc)












Could anyone provide me the details of how to determine the lag order of the distributed lags for an ADL(p,q) model in Matlab or another statistical package (and very much preferably in combination with the autoregression lags)?? Full working examples with model selection criteria (AIC and BIC) seem to be available on the Matlab website for VAR models, ARMA models etc. but I can't find one for the ADL(p,q) model. I would not have a clue to rewrite those models to ADL(p,q) myself but I have a vague feeling that such a thing would be possibe.

In the end I want to automate this analysis by first checking the lag orders p,q and then using this numbers automatically to create the regressions out of this. So basically I'm looking for a fully working example. (I want to skip part of adding and deleting regressors by hand as much as possible to get a quick idea of the distributed lags of several assets).

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.