Marketing Data Science_ Modeling Techniques In Predictive Analytics With R And Python

“Listen, I—I appreciate this whole seduction scene you’ve got going, but let me give you a tip: I’m a sure thing. OK?” —JULIA ROBERTS AS VIVIAN WARD IN Pretty Woman (1990) Mass marketing treats all customers as one group. Oneto-one marketing focuses on one customer at a time. Target marketing to selected groups of customers or market segments lies between mass marketing and oneto-one marketing. Target marketing involves directing marketing activities to those customers who are most likely to buy

About This eBook

ePUB is an open, industry-standard format for

eBooks However, support of ePUB and its many

features varies across reading devices and applications.Use your device or app settings to customize the

presentation to your liking Settings that you can

customize often include font, font size, single or doublecolumn, landscape or portrait mode, and figures thatyou can click or tap to enlarge For additional

information about the settings and features on yourreading device or app, visit the device manufacturer’sWeb site

Many titles include programming code or

configuration examples To optimize the presentation ofthese elements, view the eBook in single-column,

landscape mode and adjust the font size to the smallestsetting In addition to presenting code and

configurations in the reflowable text format, we haveincluded images of the code that mimic the presentationfound in the print book; therefore, where the reflowableformat may compromise the presentation of the codelisting, you will see a “Click here to view code image”link Click the link to view the print-fidelity code image

To return to the previous page viewed, click the Backbutton on your device or app

Marketing Data Science

Modeling Techniques in Predictive

Analytics with R and Python

T HOMAS W M ILLER

Publisher: Paul Boger

Editor-in-Chief: Amy Neidlinger

Executive Editor: Jeanne Glasser Levine

Operations Specialist: Jodi Kemper

Cover Designer: Alan Clements

Managing Editor: Kristy Hart

Manufacturing Buyer: Dan Uhrig

©2015 by Thomas W Miller

Published by Pearson Education, Inc

Old Tappan New Jersey 07675

For information about buying this title in bulk

quantities, or for special sales opportunities (which mayinclude electronic versions; custom cover designs; andcontent particular to your business, training goals,

marketing focus, or branding interests), please contactour corporate sales department at

All rights reserved No part of this book may be

reproduced, in any form or by any means, without

permission in writing from the publisher.

Printed in the United States of America

First Printing May 2015

ISBN-10: 0-13-388655-7

ISBN-13: 978-0-13-388655-9

Pearson Education LTD

Pearson Education Australia PTY, Limited

Pearson Education Singapore, Pte Ltd

Pearson Education Asia, Ltd

Pearson Education Canada, Ltd

Pearson Educación de Mexico, S.A de C.V

Pearson Education—Japan

Pearson Education Malaysia, Pte Ltd

Library of Congress Control Number: 2015937911

2 Predicting Consumer Choice

3 Targeting Current Customers

4 Finding New Customers

10 Assessing Brands and Prices

11 Utilizing Social Networks

12 Watching Competitors

13 Predicting Sales

14 Redefining Marketing Research

A Data Science Methods

A.1 Database Systems and Data PreparationA.2 Classical and Bayesian Statistics

A.3 Regression and Classification

A.4 Data Mining and Machine Learning

A.5 Data Visualization

A.6 Text and Sentiment Analysis

A.7 Time Series and Market Response Models

B Marketing Data Sources

B.9 Interviews

B.10 Focus Groups

B.11 Field Research

C Case Studies

C.1 AT&T Choice Study

C.2 Anonymous Microsoft Web DataC.3 Bank Marketing Study

C.4 Boston Housing Study

C.5 Computer Choice Study

C.6 DriveTime Sedans

C.7 Lydia E Pinkham Medicine CompanyC.8 Procter & Gamble Laundry SoapsC.9 Return of the Bobbleheads

C.10 Studenmund’s Restaurants

C.11 Sydney Transportation Study

C.12 ToutBay Begins Again

C.13 Two Month’s Salary

C.14 Wisconsin Dells

C.15 Wisconsin Lottery Sales

C.16 Wikipedia Votes

D Code and UtilitiesBibliography

Index

“Everybody loses the thing that made them It’s evenhow it’s supposed to be in nature The brave men stayand watch it happen, they don’t run.”

—QUVENZHANÉ WALLIS AS HUSHPUPPY IN Beasts of the

Southern Wild (2012)

Writers of marketing textbooks of the past would

promote “the marketing concept,” saying that marketing

is not sales or selling Rather, marketing is a matter ofunderstanding and meeting consumer needs They

would distinguish between “marketing research,” a

business discipline, and “market research,” as in

economics And marketing research would sometimes

be described as “marketing science” or “marketing

engineering.”

Ignore the academic pride and posturing of the past.Forget the linguistic arguments Marketing and sales,marketing and markets, research and science—they areone In a world transformed by information technologyand instant communication, data rule the day

Data science is the new statistics, a blending of

modeling techniques, information technology, and

business savvy Data science is also the new look of

marketing research

In introducing marketing data science, we choose topresent research about consumers, markets, and

marketing as it currently exists Research today meansgathering and analyzing data from web surfing,

crawling, scraping, online surveys, focus groups, blogsand social media Research today means finding

answers as quickly and cheaply as possible

Finding answers efficiently does not mean we mustabandon notions of scientific research, sampling, orprobabilistic inference We take care while designingmarketing measures, fitting models, describing researchfindings, and recommending actions to management.There are times, of course, when we must engage inprimary research We construct survey instruments andinterview guides We collect data from consumer

samples and focus groups This is traditional marketingresearch—custom research, tailored to the needs of eachindividual client or research question

The best way to learn about marketing data science is towork through examples This book provides a readyresource and reference guide for modeling techniques

We show programmers how to build on a foundation ofcode that works to solve real business problems

The truth about what we do is in the programs we write.The code is there for everyone to see and for some todebug To promote student learning, programs includestep-by-step comments and suggestions for taking

analyses further Data sets and computer programs are

available from the website for the Modeling Techniques

series at http://www.ftpress.com/miller/

When working on problems in marketing data science,some things are more easily accomplished with Python,others with R And there are times when it is good tooffer solutions in both languages, checking one againstthe other Together, Python and R make a strong

combination for doing data science

Most of the data in this book come from public domainsources Supporting data for many cases come from theUniversity of California–Irvine Machine Learning

Repository and the Stanford Large Network DatasetCollection I am most thankful to those who provideaccess to rich data sets for research

I have learned from my consulting work with ResearchPublishers LLC and its ToutBay division, which

promotes what can be called “data science as a service.”Academic research and models can take us only so far.Eventually, to make a difference, we need to implementour ideas and models, sharing them with one another.Many have influenced my intellectual development overthe years There were those good thinkers and goodpeople, teachers and mentors for whom I will be forevergrateful Sadly, no longer with us are Gerald Hahn

Hinkle in philosophy and Allan Lake Rice in languages

at Ursinus College, and Herbert Feigl in philosophy atthe University of Minnesota I am also most thankful toDavid J Weiss in psychometrics at the University of

Minnesota and Kelly Eakin in economics, formerly atthe University of Oregon.

Thanks to Michael L Rothschild, Neal M Ford, Peter R.Dickson, and Janet Christopher who provided

invaluable support during our years together at theUniversity of Wisconsin–Madison While serving asdirector of the A C Nielsen Center for Marketing

Research, I met the captains of the marketing researchindustry, including Arthur C Nielsen, Jr himself I metand interviewed Jack Honomichl, the industry’s

historian, and I met with Gil Churchill, first author ofwhat has long been regarded as a key textbook in

marketing research I learned about traditional

marketing research at the A C Nielsen Center for

Marketing Research, and I am most grateful for theexperience of working with its students and executiveadvisory board members Thanks go as well to Jeff

Walkowski and Neli Esipova who worked with me inexploring online surveys and focus groups when thosemethods were just starting to be used in marketingresearch

After my tenure with the University of Wisconsin–Madison, I built a consulting practice My company,Research Publishers LLC, was co-located with the

former Chamberlain Research Consultants SharonChamberlain gave me a home base and place to practicethe craft of marketing research It was there that initialconcepts for this book emerged:

What could be more important to a business than understanding its customers, competitors, and markets? Managers need a

coherent view of things With consumer research, product

management, competitive intelligence, customer support, and

management information systems housed within separate

departments, managers struggle to find the information they

need Integration of research and information functions makes

more sense (Miller 2008).

My current home is the Northwestern University School

of Professional Studies I support courses in three

graduate programs: Master of Science in Predictive

Analytics, Advanced Certificate in Data Science, andMaster of Arts in Sports Administration Courses inmarketing analytics, database systems and data

preparation, web and network data science, and datavisualization provide inspiration for this book

I expect Northwestern’s graduate programs to prosper

as they forge into new areas, including analytics

entrepreneurship and sports analytics Thanks to

colleagues and staff who administer these exceptionalgraduate programs, and thanks to the many studentsand fellow faculty from whom I have learned

Amy Hendrickson of TEXnology Inc applied her craft,making words, tables, and figures look beautiful in print

—another victory for open source Lorena Martin

reviewed the book and provided much needed feedback.Roy Sanford provided advice on statistical explanations.Candice Bradley served dual roles as a reviewer and

copyeditor for all books in the Modeling Techniques

series I am grateful for their guidance and

encouragement

Thanks go to my editor, Jeanne Glasser Levine, andpublisher, Pearson/FT Press, for making this and other

books in the Modeling Techniques series possible Any

writing issues, errors, or items of unfinished business,

of course, are my responsibility alone

My good friend Brittney and her daughter Janiya keep

me company when time permits And my son Daniel isthere for me in good times and bad, a friend for life Mygreatest debt is to them because they believe in me.Thomas W Miller

Glendale, California

April 2015

2.2 Correlation Heat Map for Explanatory Variables inthe Sydney Transportation Study

2.3 Logistic Regression Density Lattice

2.4 Using Logistic Regression to Evaluate the Effect ofPrice Changes

3.1 Age and Response to Bank Offer

3.2 Education Level and Response to Bank Offer

3.3 Job Type and Response to Bank Offer

3.4 Marital Status and Response to Bank Offer

3.5 Housing Loans and Response to Bank Offer

3.6 Logistic Regression for Target Marketing (DensityLattice)

3.7 Logistic Regression for Target Marketing

4.2 Response to Term Deposit Offers by Market

5.6 Logistic Regression for the Probability of Switching(Confusion Mosaic)

5.7 A Classification Tree for Predicting Consumer

Choices about Service Providers

5.8 Logistic Regression for Predicting Customer

Retention (ROC Curve)

5.9 Nạve Bayes Classification for Predicting CustomerRetention (ROC Curve)

5.10 Support Vector Machines for Predicting CustomerRetention (ROC Curve)

6.1 A Product Similarity Ranking Task

6.2 Rendering Similarity Judgments as a Matrix

6.3 Turning a Matrix of Dissimilarities into a

Perceptual Map

6.4 Indices of Similarity and Dissimilarity between

Pairs of Binary Variables

6.5 Map of Wisconsin Dells Activities Produced by

8.1 Dodgers Attendance by Day of Week

8.2 Dodgers Attendance by Month

8.3 Dodgers Weather, Fireworks, and Attendance

8.4 Dodgers Attendance by Visiting Team

8.5 Regression Model Performance: Bobbleheads andAttendance

9.1 Market Basket Prevalence of Initial Grocery Items9.2 Market Basket Prevalence of Grocery Items by

Category

9.3 Market Basket Association Rules: Scatter Plot

9.4 Market Basket Association Rules: Matrix BubbleChart

9.5 Association Rules for a Local Farmer: A NetworkDiagram

10.1 Computer Choice Study: A Mosaic of Top Brandsand Most Valued Attributes

10.2 Framework for Describing Consumer Preferenceand Choice

10.3 Ternary Plot of Consumer Preference and Choice10.4 Comparing Consumers with Differing Brand

11.5 Degree Distributions for Network Models

11.6 Network Modeling Techniques

12.1 Competitive Intelligence: Spirit Airlines Flying High13.1 Scatter Plot Matrix for Restaurant Sales and

14.1 Competitive Analysis for the Custom ResearchProvider

14.2 A Model for Strategic Planning

14.3 Data Sources in the Information Supply Chain14.4 Client Information Sources and the World WideWeb

14.5 Networks of Research Providers, Clients, and

Intermediaries

A.1 Evaluating the Predictive Accuracy of a BinaryClassifier

A.2 Linguistic Foundations of Text Analytics

A.3 Creating a Terms-by-Documents Matrix

B.1 A Framework for Marketing Measurement

B.2 Hypothetical Multitrait-Multimethod Matrix

B.3 Framework for Automated Data Acquisition

B.4 Demographic variables from Mintel survey

B.5 Sample questions from Mintel movie-going surveyB.6 Open-Ended Questions

B.7 Guided Open-Ended Question

B.8 Behavior Check List

B.9 From Check List to Click List

B.10 Adjective Check List

B.11 Binary Response Questions

B.12 Rating Scale for Importance

B.13 Rating Scale for Agreement/Disagreement

B.24 Paired Comparison Choice Task

B.25 Choice Set with Three Product Profiles

B.26 Menu-based Choice Task

B.27 Elimination Pick List

B.28 Factors affecting the validity of experimentsB.29 Interview Guide

B.30 Interview Projective Task

C.1 Computer Choice Study: One Choice Set

7.1 Analysis of Deviance for New Product Field Test:Procter & Gamble Laundry Soaps

8.1 Bobbleheads and Dodger Dogs

8.2 Regression of Attendance on Month, Day of Week,and Bobblehead Promotion

9.1 Market Basket for One Shopping Trip

9.2 Association Rules for a Local Farmer

10.1 Contingency Table of Top-ranked Brands and MostValued Attributes

10.2 Market Simulation: Choice Set Input

10.3 Market Simulation: Preference Shares in a

Hypothetical Four-brand Market

12.1 Competitive Intelligence Sources for Spirit Airlines

13.1 Fitted Regression Model for Restaurant Sales

13.2 Predicting Sales for New Restaurant Sites

A.1 Three Generalized Linear Models

B.1 Levels of measurement

C.1 Variables for the AT&T Choice Study

C.2 Bank Marketing Study Variables

C.3 Boston Housing Study Variables

C.4 Computer Choice Study: Product Attributes

C.5 Computer Choice Study: Data for One IndividualC.6 Hypothetical profits from model-guided vehicleselection

C.7 DriveTime Data for Sedans

C.8 DriveTime Sedan Color Map with Frequency

Counts

C.9 Variables for the Laundry Soap Experiment

C.10 Cross-Classified Categorical Data for the LaundrySoap Experiment

C.11 Variables for Studenmund’s Restaurants

C.12 Data for Studenmund’s Restaurants

C.13 Variables for the Sydney Transportation StudyC.14 ToutBay Begins: Website Data

C.15 Diamonds Data: Variable Names and Coding RulesC.16 Dells Survey Data: Visitor Characteristics

C.17 Dells Survey Data: Visitor Activities

C.18 Wisconsin Lottery Data

C.19 Wisconsin Casino Data

C.20 Wisconsin ZIP Code Data

C.21 Top Sites on the Web, September 2014

1.1 Measuring and Modeling Individual Preferences (R)1.2 Measuring and Modeling Individual Preferences(Python)

2.1 Predicting Commuter Transportation Choices (R)2.2 Predicting Commuter Transportation Choices

(Python)

3.1 Identifying Customer Targets (R)

4.1 Identifying Consumer Segments (R)

4.2 Identifying Consumer Segments (Python)

5.1 Predicting Customer Retention (R)

6.1 Product Positioning of Movies (R)

6.2 Product Positioning of Movies (Python)

6.3 Multidimensional Scaling Demonstration: US Cities(R)

6.4 Multidimensional Scaling Demonstration: US Cities(Python)

6.5 Using Activities Market Baskets for Product

Positioning (R)

6.6 Using Activities Market Baskets for Product

Positioning (Python)

6.7 Hierarchical Clustering of Activities (R)

7.1 Analysis for a Field Test of Laundry Soaps (R)

8.1 Shaking Our Bobbleheads Yes and No (R)

8.2 Shaking Our Bobbleheads Yes and No (Python)9.1 Market Basket Analysis of Grocery Store Data (R)9.2 Market Basket Analysis of Grocery Store Data(Python to R)

10.1 Training and Testing a Hierarchical Bayes Model(R)

10.2 Analyzing Consumer Preferences and Building aMarket Simulation (R)

11.1 Network Models and Measures (R)

11.2 Analysis of Agent-Based Simulation (R)

11.3 Defining and Visualizing a Small-World Network(Python)

11.4 Analysis of Agent-Based Simulation (Python)12.1 Competitive Intelligence: Spirit Airlines FinancialDossier (R)

13.1 Restaurant Site Selection (R)

13.2 Restaurant Site Selection (Python)

D.1 Conjoint Analysis Spine Chart (R)

D.2 Market Simulation Utilities (R)

D.3 Split-plotting Utilities (R)

D.4 Utilities for Spatial Data Analysis (R)

D.5 Correlation Heat Map Utility (R)

D.6 Evaluating Predictive Accuracy of a Binary

Classifier (Python)

1 Understanding Markets

“What makes the elephant guard his tusk in the mistymist, or the dusky dusk? What makes a muskrat guardhis musk?”

—BERT LAHR AS COWARDLY LION IN The Wizard of Oz

(1939)

While working on the first book in the Modeling

Techniques series, I moved from Madison, Wisconsin to

Los Angeles I had a difficult decision to make aboutmobile communications I had been a customer of U.S.Cellular for many years I had one smartphone and twodata modems (a 3G and a 4G) and was quite satisfiedwith U.S Cellular services In May of 2013, the companyhad no retail presence in Los Angeles and no 4G service

in California Being a data scientist in need of an

example of preference and choice, I decided to assess

my feelings about mobile phone services in the Los

Angeles market

The attributes in my demonstration study were the

mobile provider or brand, startup and monthly costs, ifthe provider offered 4G services in the area, whether theprovider had a retail location nearby, and whether theprovider supported Apple, Samsung, or Nexus phones inaddition to tablet computers Product profiles,

representing combinations of these attributes, were

easily generated by computer My consideration setincluded AT&T, T-Mobile, U.S Cellular, and Verizon Igenerated sixteen product profiles and presented them

to myself in a random order Product profiles, theirattributes, and my ranks, are shown in table 1.1

Table 1.1 Preference Data for Mobile

Communication Services

A linear model fit to preference rankings is an example

of traditional conjoint analysis, a modeling technique

designed to show how product attributes affect

purchasing decisions Conjoint analysis is really

conjoint measurement Marketing analysts present

product profiles to consumers Product profiles are

defined by their attributes By ranking, rating, or

choosing products, consumers reveal their preferencesfor products and the corresponding attributes that

define products The computed attribute importancevalues and part-worths associated with levels of

attributes represent measurements that are obtained as

a group or jointly—thus the name conjoint analysis The

task—ranking, rating, or choosing—can take many

forms

When doing conjoint analysis, we utilize sum contrasts,

so that the sum of the fitted regression coefficients

across the levels of each attribute is zero The fittedregression coefficients represent conjoint measures of

utility called part-worths Part-worths reflect the

strength of individual consumer preferences for eachlevel of each attribute in the study Positive part-worthsadd to a product’s value in the mind of the consumer.Negative part-worths subtract from that value When wesum across the part-worths of a product, we obtain ameasure of the utility or benefit to the consumer

To display the results of the conjoint analysis, we use a

special type of dot plot called the spine chart, shown in

figure 1.1 In the spine chart, part-worths can be

displayed on a common, standardized scale across

attributes The vertical line in the center, the spine, isanchored at zero

Figure 1.1 Spine Chart of Preferences for Mobile

Communication Services

The part-worth of each level of each attribute is

displayed as a dot with a connecting horizontal line,extending from the spine Preferred product or servicecharacteristics have positive part-worths and fall to theright of the spine Less preferred product or servicecharacteristics fall to the left of the spine

The spine chart shows standardized part-worths andattribute importance values The relative importance ofattributes in a conjoint analysis is defined using theranges of part-worths within attributes These

importance values are scaled so that the sum across allattributes is 100 percent Conjoint analysis is a

measurement technology Part-worths and attributeimportance values are conjoint measures

What does the spine chart say about this consumer’spreferences? It shows that monthly cost is of

considerable importance Next in order of importance is4G availability Start-up cost, being a one-time cost, is

much less important than monthly cost This consumerranks the four service providers about equally And

having a nearby retail store is not an advantage Thisconsumer is probably an Android user because we seehigher importance for service providers that offer

Samsung phones and tablets first and Nexus second,while the availability of Apple phones and tablets is oflittle importance

This simple study reveals a lot about the consumer—itmeasures consumer preferences Furthermore, the

linear model fit to conjoint rankings can be used to

predict what the consumer is likely to do about mobilecommunications in the future

Traditional conjoint analysis represents a modelingtechnique in predictive analytics Working with groups

of consumers, we fit a linear model to each individual’sratings or rankings, thus measuring the utility or part-worth of each level of each attribute, as well as the

relative importance of attributes

The measures we obtain from conjoint studies may beanalyzed to identify consumer segments Conjoint

measures can be used to predict each individual’s

choices in the marketplace Furthermore, using conjointmeasures, we can perform marketplace simulations,exploring alternative product designs and pricing

policies Consumers reveal their preferences in

responses to surveys and ultimately in choices theymake in the marketplace

Marketing data science, a specialization of predictiveanalytics or data science, involves building models ofseller and buyer preferences and using those models tomake predictions about future marketplace behavior.Most of the examples in this book concern consumers,but the ways we conduct research—data preparation andorganization, measurements, and models—are relevant

to all markets, consumer and business markets alike

business-to-Managers often ask about what drives buyer choice.They want to know what is important to choice or whichfactors determine choice To the extent that buyer

behavior is affected by product features, brand, andprice, managers are able to influence buyer behavior,increasing demand, revenue, and profitability

Product features, brands, and prices are part of the

mobile phone choice problem in this chapter But thereare many other factors affecting buyer behavior—

unmeasured factors and factors outside managementcontrol Figure 1.2 provides a framework for

understanding marketplace behavior—the choices ofbuyers and sellers in a market

Figure 1.2 The Market: A Meeting Place for Buyers

and Sellers

A market, as we know from economics, is the locationwhere or channel through which buyers and sellers gettogether Buyers represent the demand side, and sellersthe supply side To predict what will happen in a market

—products to be sold and purchased, and the clearing prices of those products—we assume that

market-sellers are profit-maximizers, and we study the pastbehavior and characteristics of buyers and sellers We

build models of market response This is the job of

marketing data science as we present it in this book

Ask buyers what they want, and they may say, the best

of everything Ask them what they would like to spend, and they may say, as little as possible There are

limitations to assessing buyer willingness to pay andproduct preferences with direct-response rating scales,

or what are sometimes called self-explicative scales.Simple rating scale items arranged as they often are,with separate questions about product attributes,

brands, and prices, fail to capture tradeoffs that are

fundamental to consumer choice To learn more frombuyer surveys, we provide a context for responding andthen gather as much information as we can This is whatconjoint and choice studies do, and many of them do itquite well In the appendix B (pages 312 to 337) we

provide examples of consumer surveys of preferenceand choice

Conjoint measurement, a critical tool of marketing datascience, focuses on buyers or the demand side of

markets The method was originally developed by Luceand Tukey (1964) A comprehensive review of conjointmethods, including traditional conjoint analysis, choice-based conjoint, best-worst scaling, and menu-basedchoice, is provided by Bryan Orme (2013) Primary

applications of conjoint analysis fall under the headings

of new product design and pricing research, which wediscuss later in this book

Exhibits 1.1 and 1.2 show R and Python programs foranalyzing ranking or rating data for consumer

preferences The programs perform traditional conjointanalysis The spine chart is a customized data

visualization for conjoint and choice studies We showthe R code for making spine charts in appendix D,

exhibit D.1 starting on page 400 Using standard R

graphics, we build this chart one point, line, and textstring at a time The precise placement of points, lines,and text is under our control

Exhibit 1.1 Measuring and Modeling Individual

Preferences (R)

Click here t o v iew code image

# Traditional Conjoint Analysis (R)

# R preliminaries to get the user-defined

function for spine chart:

# place the spine chart code file

<R_utility_program_1.R>

# in your working directory and execute it by

# source("R_utility_program_1.R")

# Or if you have the R binary file in your

working directory, use

# load(file="mtpa_spine_chart.Rdata")

# spine chart accommodates up to 45

part-worths on one page

# |part-worth| <= 40 can be plotted directly

on the spine chart

# |part-worths| > 40 can be accommodated

retail = c("Retail NO","Retail YES"),

apple = c("Apple NO","Apple YES"),

samsung = c("Samsung NO","Samsung YES"),

google = c("Nexus NO","Nexus YES")),

nalternatives = 1, nblocks=1, seed=9999)

print(questionnaire(provider.survey)) # print survey design for review

sink("questions_for_survey.txt") # send survey

to external text file

questionnaire(provider.survey)

sink() # send output back to the screen

# user-defined function for plotting

descriptive attribute names

effect.name.map <- function(effect.name) {

if(effect.name=="brand") return("Mobile

Service Provider")

if(effect.name=="startup") return("Start-up Cost")

if(effect.name=="monthly") return("Monthly Cost")

if(effect.name=="service") return("Offers 4G Service")

if(effect.name=="retail") return("Has Nearby Retail Store")

if(effect.name=="apple") return("Sells Apple Products")

if(effect.name=="samsung") return("Sells Samsung Products")

# main effects model specification

main.effects.model <- {ranking ~ brand +

startup + monthly + service +

retail + apple + samsung + google}