Many times they tell us about the typical prediction models based mainly on linear econometric models that try to explain a variable (explained variable) as a function of another (explanatory variable), such as the relationship between income and consumption of a person. But there are occasions in which none of the linear models in the parameters (linear regression, polynomial regression, exponential, etc.) is useful to make a forecast. For example, tourist flows in a tourist destination are determined by high and low season, so a continuous time series will probably give a non-representative R ^ 2 for the model.
Advertisements
The technique is relatively simple, you must add a series of variables that explain seasonality, with a matrix very similar to the matrix of identity, in this way the regression formula will have a series of zeros that will cancel those values and that will consider the ones that cross the matrix of identity. Well, here is an example in excel of a case in which there is a seasonal demand for a series of 24 months so that you can take it as an example.
Download here
Seasonal-linear-regression
Advertisements
I hope it will be you useful….
