Chapter 14 Introduction

393

CHAPTER 14

Regression

Simple Linear Regression

Prediction versus Relation

Regression with z Scores

Determining the Regression Equation

The Standardized Regression Coefficient and Hypothesis Testing with Regression

Interpretation and Prediction

Regression and Error

Applying the Lessons of Correlation to Regression

Regression to the Mean

Proportionate Reduction in Error

Multiple Regression

Understanding the Equation

Multiple Regression in Everyday Life

BEFORE YOU GO ON

  • You should understand the six steps of hypothesis testing (Chapter 7).

  • You should understand the concept of effect size (Chapter 8).

  • You should understand the concept of correlation (Chapter 13).

  • You should be able to explain the limitations of correlation (Chapter 13).

394

image
Facebook and Social Capital The prediction tools introduced in this chapter helped researchers determine that increased use of Facebook predicts higher levels of social capital.
© hamideddine Bouali/Demotix/Demotix/Demotix/Corbis

In 2004, college student Mark Zuckerberg created the social networking site Facebook, which soon exploded in popularity across college campuses. By 2012, Facebook had raised $16 billion in a public stock offering. As of 2015, it has more than 1.4 billion users. As Facebook use ballooned, researchers at Michigan State University (MSU) (Ellison, Steinfeld, & Lampe, 2007) wanted to understand what college students were getting out of their Facebook relationships—that is, what they were gaining in social capital.

One type of social capital the researchers studied was “bridging” social capital: the loose social connections we think of as acquaintances rather than friends. The researchers hypothesized that greater use of Facebook would predict more bridging social capital. They measured bridging social capital by asking students to rate several items, such as “I feel I am part of the MSU community” and “At MSU, I come into contact with new people all the time.”

Obviously, the amount of time spent on Facebook is only one of many influences that determine how much social capital students get from Facebook. Clear answers to the research question are probably complicated by many different factors such as shyness, gender, and quality of Internet connections. The MSU Facebook study had to control for the influence of many variables, but the data still supported the researcher’s hypothesis: the more students used Facebook, the higher they tended to score on a measure of social capital. But how did the researchers isolate the effect of time on Facebook from other variables?

The analytical methods we learn in this chapter build on correlation to help us to create prediction tools. We begin by learning how to use just one scale variable to predict the outcome on a second scale variable. Then we discuss the limitations of this method—limitations that are similar to those we encountered with correlation. Finally, we expand this analytical method to use multiple scale variables to predict the outcome on another scale variable.