khakieconomics/demos

Demonstrations of various techniques

A collection of demonstrations for blog posts, online arguments, etc.

prox_weighting: Using proximity score weighting as a method of modelling model changes

This script goes with my blog post on 4 characteristics of furute macroeconometric modelling.

ads_demo: Using a vector autoregression to work out the impact of advertising spend

The script in this repo gives a simple demonstration of how one could do some analysis of advertising expenditure using a vector autoregression model. The model is fairly general, but makes the following asumptions:

• Both ad placements and website traffic are mutually endogenous
• There are no trends or unit roots. If your data do have trends and/or unit roots, look at library(urca) and the VECM function in vars.
• I have not included any exogenous regressors. In general, you would have day of the week effects, time of day effects, etc. You can include a matrix of these in the call to VARS. You'll want to drop the intercept term "const" if you have these effects (colinearity).
• The most important thing is that it assumes that your treatment is conditionally exogenous. That is, you are trying to estimate the causal impact of advertising expenditure on website traffic. If you don't have a conditionally exogenous treatment, then you are only looking at average comovements, which mean a whole bunch less.