Two earlier posts (here and here) described a simple hierarchical linear model (HLM) for Black Friday Retail Sales using data from an article in the Washington Post (here). The pedagogical purpose of the HLM exercise was to display one answer to Simpson's Paradox. It wasn't meant to be a general recommendation for the analysis of retail sales.
The analysis of aggregate retail sales is essentially a time series problem in that we are analyzing sales over time for, ideally, multiple years. HLMs can be written to handle time series problems, a topic that I will return to in a later post. The analysis of aggregate US retail sales, however, can be analyzed with a straight-forward state space time series model. If you are interested in the pure time series approach, I have done that in another post (here).
The insight from the time series analysis is that US Retail Sales are being driven by the World economy which makes sense in a world of globalized retail trade. The idea that Black Friday Sales might be a good predictor of aggregate retail sales is, based on purely theoretical considerations, not a very appealing hypothesis.
No comments:
Post a Comment