If you combine effect coding (or, as you’ll see, dummy coding) with linear regression, you can get the forecasts of a seasonal time series in one step. Again, you could use Holt’s method instead of regression. (It would have been feasible to use Holt’s method in Figure 5.7 instead of linear regression.)įigure 5.14 shows how you can use linear regression just as is done in Figure 5.7, to forecast from deseasonalized observations prior to reseasonalizing them using the seasonal indexes. Figure 5.7 shows how the same method uses regression to forecast from deseasonalized quarters before adding seasonal effects back in. Figure 5.5 shows how the simple averages method uses regression to calculate a year-to-year trend and distribute that trend equally across the quarters. The two broad methods that this chapter has already discussed, those of simple averages and of moving averages, also employ linear regression. Effect coding is just a way of representing which season a particular observation belongs to, and Excel makes it particularly easy to set up. Learn More Buy Linear Regression with Coded VectorsĪnother method for dealing with seasonal time series, whether trended or not, employs linear regression in combination with effect coding. More Predictive Analytics: Microsoft Excel
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |