Weighted Linear Regression
Bubble charts are a natural way to visualize many types of summarized data. The data is located on the chart according to X- and Y-values, and the size of each bubble is the frequency count. Alas, when Excel calculates the best-fit line in a bubble chart, it does not take into account the sizes of bubbles. The resulting best-fit line does a poor job showing trends in the data.
The way to solve this problem is by using weighted linear regression to take the sizes of the bubbles into account. Unfortunately, this capability is not built into Excel directly. There are two ways to do the calculation. One is to apply the formulas from the previous section, adjusting the various intermediate sums for the frequencies. The other uses special...