I have daily TV advertising, sales data, and website visitor data for a product and would like to estimate the effect (drag) of weekend advertising on Monday sales. The hypothesis is that consumers view the commercials on TV during the weekend and then wait until Monday (when they are bored at work, on their computers) in order to go online and purchase the product. I have 2+ years worth of data. There is obviously seasonality throughout the year and seasonality throughout the week. Using a simple lag-model is not sufficient because it will seek to estimate sales at time "t" based on spend from t, t-1, t-2, etc. In my case, I am concerned with the lag effect for every time t, but moreso for the effect of Sunday and Saturday spend on Monday sales. I am not so much concerned with the effect of Wednesday and Thursday spend on Friday spend. What type of model will allow me to include a term specific to the effect of Sunday and Saturday spend on Monday orders?
Hi Garrett,
You can simply generate a dummy variable for Saturday and Sunday (together or separately, depending on whether you believe they jointly impact Monday, or if they operate separately), and then use that covariate in your model. You should be able to determine if that dummy has a lagged effect on Monday's sales.
Ariel
Awesome! This idea came to me when I stepped away from the data for a bit yesterday. My first model showed some promise and I am now building on it!
- Badges
- Science topic
- Similar topics
- Psychology