Microsoft Office Tutorials and References

In Depth Information

**Considerations When Using Regression Analysis**

lem is that this line has no predictive ability. If you base your future sales

on this line, you will get results that will vary greatly from the prediction.

Figure 14.13.

Figure 14.13.
This data set has no correlation to time.

happily pre-

dicts a line, but it is severely wrong most of the time.

This data set has no correlation to time. LINEST

LINEST happily pre-

dicts a line, but it is severely wrong most of the time.

Part of the results of regression analysis are statistics that tell how

well the regression line fits the actual data. You should always check

statistics such as r-squared or the standard error to see if the past data

shows a relationship between the variables. The r-squared value is a value

between 0 and 1. The closer that r-squared is to 1, the better the regression