TY - JOUR
T1 - A Primer on Using Behavioral Data for Testing Theories in Advertising Research
AU - Liu-Thompkins, Yuping
AU - Malthouse, Edward C.
N1 - Publisher Copyright:
Copyright © 2017, American Academy of Advertising.
PY - 2017/1/2
Y1 - 2017/1/2
N2 - Interactions with and between customers in digital, social, and mobile environments are commonly recorded, producing behavioral data that have the potential to advance advertising research. This article provides an accessible guide on how to leverage such data for advertising researchers who may have thus far relied mostly on lab experiment or survey data. Specifically, we suggest potential sources for behavioral data and present a process for analyzing and interpreting behavioral data. Each step of the process is discussed: exploring, understanding and preparing data; specifying and estimating models; and interpreting and presenting the results. Some fundamental issues with using multiple regression to analyze such data are covered, including standardization, outliers, transformations, multicollinearity, and the omitted variable bias. We also discuss issues that are especially problematic with using behavioral data in advertising research, including endogeneity, count data, data with many zeros, and grouped data. More advanced versions of regression that address these issues are surveyed, including instrumental variables, propensity scoring, generalized linear models, and mixed models. General advice for thinking about behavioral data is provided.
AB - Interactions with and between customers in digital, social, and mobile environments are commonly recorded, producing behavioral data that have the potential to advance advertising research. This article provides an accessible guide on how to leverage such data for advertising researchers who may have thus far relied mostly on lab experiment or survey data. Specifically, we suggest potential sources for behavioral data and present a process for analyzing and interpreting behavioral data. Each step of the process is discussed: exploring, understanding and preparing data; specifying and estimating models; and interpreting and presenting the results. Some fundamental issues with using multiple regression to analyze such data are covered, including standardization, outliers, transformations, multicollinearity, and the omitted variable bias. We also discuss issues that are especially problematic with using behavioral data in advertising research, including endogeneity, count data, data with many zeros, and grouped data. More advanced versions of regression that address these issues are surveyed, including instrumental variables, propensity scoring, generalized linear models, and mixed models. General advice for thinking about behavioral data is provided.
UR - http://www.scopus.com/inward/record.url?scp=85010670180&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85010670180&partnerID=8YFLogxK
U2 - 10.1080/00913367.2016.1252289
DO - 10.1080/00913367.2016.1252289
M3 - Article
AN - SCOPUS:85010670180
SN - 0091-3367
VL - 46
SP - 213
EP - 225
JO - Journal of Advertising
JF - Journal of Advertising
IS - 1
ER -