Statistical Analysis with R For Dummies

Chapter 16

Curvilinear Regression: When Relationships Get Complicated

IN THIS CHAPTER

Understanding exponents

Connecting logarithms to regression

Pursuing polynomials

In Chapters 14 and 15, I describe linear regression and correlation — two concepts that depend on the straight line as the best-fitting summary of a scatterplot.

But a line is isn’t always the best fit. Processes in a variety of areas, from biology to business, conform more to curves than to lines.

For example, think about when you learned a skill — like tying your shoelaces. When you first tried it, it took quite a while didn’t it? And then whenever you tried it again, it took progressively less time for you to finish, right? Until finally, you can tie your shoelaces very quickly but you can’t really get any faster — you’re now doing it is as efficiently as you can.

If you plotted shoelace-tying-time (in seconds) on the y-axis and trials (occasions when you tried to tie your shoes) on the x-axis, the graph might look something like Figure 16-1. A straight line is clearly not the best summary of a plot like this.

FIGURE ...