I am working on the revision to a manuscript in which I used Google Trends data to evaluate whether a particular event affected Google search patterns. I used an ARIMA model to forecast the expected search pattern had the event not occurred. The peer reviewer remarked that our study design did not control enough for confounding and asked that I provide the average accuracy of ARIMA projections in other time frames (previous year, previous months). I was hoping someone on this discussion board would be able to tell me the best way to address this comment. I appreciate your help.
closed as unclear what you're asking by mkt, Michael Chernick, mdewey, BruceET, Siong Thye Goh Sep 18 at 17:40
Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.
Some measure of out-of-sample forecast accuracy would be good. The main idea is to compare the mean square error of the ARIMA model with a benchmark model. A good benchmark is the mean of the series up to time $t$. This is also the idea of the Campbell Thomson (2009) out-of-sample $R^2$, which also has the advantage that it can interpreted as an $R^2$ measure for forecasts.
The proper thing to do is to refer you to the original paper, but I couldn't find a link. However, the way to calculate it (which is fairly easy) is described in this working paper:
Hope it helps.