Do you want to predict the future? This simple regression calculation is the most commonly used to forecast the future and other linear outcomes. If your company had sales over a period of several years you can plot those points on a graph. If those points were to fit neatly on a straight line, then you would have no need for regression. You could predict the next years sales easily. However, that is not the real world. It is far more likely that the plotted points will be much more random. How do you mathematically define what the trend of the random points are? Least squares regression is the answer.
The least squares regression calculation minimizes the error of the constructed line. Every data point contributes to the line. The result is a best fit line for the data given. The equations for this calculation can be found in any statistics textbook. The underlying assumption is that the data is linear.
The following example illustrates how linear regression is calculated. The spreadsheet is interactive, so feel free to play with it.
It is instructive to change data points to see how the equation and the graphs change. If you want to use your own data, please do. The calculations
will be accurate for any number of data points, if you take the following precautions:
If you have fewer than 25 data points, there are 3 things you must be aware of:
1. You must zero out the data points that you are not using at the end of the data list.
2. You must enter the correct number of data points in the appropriate cell.
3. The graphs will only work correctly with 25 data points.
This simple regression calculation illustrates a key method for predicting or forecasting the future. It is a very powerful tool, if used correctly. Using it correctly means that you must be able to interpret the meaning of the parameters of slope, intercept and correlation. You will find the interpretation of those parameters here.