There are two ways I can think of:
With the following function, one can retrieve the linear regression coefficients of two sets of data:
LINEST(y-axis,x-axis) ==> y=ax+b
The second set of data is the accompanying x-axis. This can be a normal numbering, as shown in the first sheet of the example file.
Once the regression factors are known, one can extrapolate any unknown x or y beyond the known set of data:
- known x: ax+b
- known y: (-b+y)/x
Since R2=1, the numbers given are exact values. I formated the values to be rounded in the example, because the calculation shows floating point errors.
Google has a way of predicting the outcome of a serie, based on that serie. If you drag the serie like shown in the screenshot, then the outcome will appear automatically:
I've created an example file for you: Extrapolate
A more practical example: http://webapps.stackexchange.com/a/35668/29140
The regression method will give you always a value, based on the factors. The more un-linear the sets of data are (R2<1), the more predicted they are.