Let's say that I have data depicting the number of museum visits per day. My challenge is to understand how certain external (exogenous?) variables, e.g., weather and advertizing affect the number of daily visits. Moreover, I need to predict how the number of visits will change when the external variables are changed.

My data has some quite obvious seasonal fluctuations where weekends and certain holidays attract more visitors. A yearly seasonal length can also be assumed to exist.

It has been quite some time since I last worked with time series, and need therefore to be pointed in the right direction. What is the approach I should use to model this data? Pre-processing steps? Model type? Typical pitfalls? Does anybody know of worked examples I could get inspiration from?


You could use a regression model with ARMA errors. Or more generally, a transfer function model.

See http://otexts.com/fpp/9/1/ for an introduction to regression with ARMA errors.

On transfer function models, Shumway and Stoffer (2011) provide a good intro.

  • $\begingroup$ Thank you. I'm studying the material and topics as we speak. However, one question: Let's say that I have found a ARMA-model that iI'm happy with. Will I be able to deduct how changes in the independent varialbes will affect the outcome (numer of visitors)? That is, similarly to say, what a normal regression model would. $\endgroup$ – Figaro Mar 19 '13 at 14:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.