Many people’s instincts when building a marketing mix model is to include as many control variables as possible. This often seems intuitively correct since many people are taught that “more controls is better” when they learn about regression analysis, but unfortunately the approach is often wrong, and in this article we'll explore why – including with some code examples you can find here. Every analyst learns about omitted variable bias early on: if you leave an important variable out of your analysis, your estimates will be off. The lesson sticks, but what doesn’t get talked about enough is the problem of included variable bias. But including the wrong variable(s) in a model can bias your estimates just as badly as…