WidgeCorp The idea behind using a regression model for forecast sales for cold beverages next year is to help the company make a better decision about that industry. The way this is done is to test all the possible influences on cold beverage demand, and apply the results. There are two different approaches we can take -- the simple approach and the complex approach. Which approach is better depends on what the decision we are making is. Obviously there are differences between the amount of time, effort and most importantly money that go into the two approaches.

The simple approach would be useful here if the decision is simply yes/no, should we enter the cold beverage market? In that case, precise demand statistics are probably not required. We can use next year's sales as the dependent variable, and then have a handful of independent variables for the past several years including cold beverage sales, population, median income and weighted...

The relationships between a handful of variables will give us a rough estimate of how the industry is expected to perform next year, given expectations from demographers and climate modelers. If we have a general idea of how fast the cold beverage industry is rising, that is going to help us decide if it is a good industry to be in.
A more complex form of analysis would be conducted the same way, but with more variables. This would allow us to fine-tune our decision making. Suppose we have already decided to enter the industry, but we had to make decisions with limited resources, so we can only enter five states. We want to know the five with the best growth. We might want to know what states have the highest sales from independent cold beverage producers (not Coke or Pepsi). With more complex analysis a much wider range…