Imtroducion al

~ Ca su relatianship
.
FuMOtAnS Of MR : Desoriptiom , Epplamation Prediotion ,

.
MR -
process f goalBatatlethodImterpret t relatomshipt
1 deoisions

imtormationknowledge
→
data decisons
→

.
MB : planning , collection , analys3 , commumication (scentifio process ]
时

I assigned ireuBed )
(
.

“
fnd
newidentify
-
mormation needs Background
searcy
defme problems & questons mitial → redene)
弘
specify objeotnes 0 Mformation value
v
, Prionary =
Survey Experiment, ObSerlaton
,

Secondary Database , Big Data

:

3 . Collect & Analyre: T - test , Corvelation , RegreSSTon , ANOUA

Sampling
x .

Sample 特 :
UHbidsed vepresentative sirablo reacheable
, , ,
data
Populaton censusrall
o

)
} 、
2 r
.

must be def ⼂med

~ -
.

Errors M 3 amplmg ? 时
x by Mterviewe 。
e= M - e, g .

samplingernrrandostematn) canavoid . nardtoquantifyhes ⼀

3 # estmated & controlled→ random & ⼤
/ ampingerrorNon
'
m…

)
gampling plan
.

i
step 3 sampling frame : source materialor devine from whioh a sample
step 4 method 13 drawn
sampling
sample sire

simple ran dom table , Pi population

1
sire
Probab7 ity sampling

1
部 size
randommoun
probablity of Gystematio SKEP INTERVAL = ssize

randomeng
n samplelideaL
straitified MECE e. 9 ,
age
Gluster representatNe eg 地理 avea “

Conlenience -

i
nonprobaBTity sampling
Judgerment
probabnity of seleotng
quota based on demographios or Elassificatnn selected by researcher
Bunkmown 1 eaby
了 mowball Z202
n=
step 可 Samplesize E
↑
Cost f acouraay msignificantly )2 budget31 Rules of
I. ) thumb
出 Traditional stat 13ti 0 .forpobabnitysample only , MoSt Scientifn ) ( Mormdl dBfributon 5 , ,
error,
E 2

level confidenceJ
Survey Method

Research design : 7 valid conolusion 时

systemat1planning
.
r

1. Person i tramed numan mterviewer) In Intercepe

- Home Executve Mall terceptpurohase
2 - 2
Conduct:
Speed of data collection: Fast 2
Pros:
Cost: Moderate to high
Immediate feedback; (go beyond responses) telephone :

|
Geographical flexibility: Confined, urban bias (quality concern)
Better explanation about the task, survey, or questionnaire
Respondent cooperation: Moderate to low (quality concern)
Design: Flexibility in survey design (showing products/ food tests/ etc.) traditional
Questionnaire length: Moderate to long • Cons:
Item non-response: Medium um"
• Interviewer error -> receive biased responses
Special features: Taste test, viewing of TV commercials possible
Quality:
• Expensive and time consuming ( GomButer)
Possibility of respondent misunderstanding: Lowest CATS
Degree of interviewer influence of answers: Highest Anonymity of respondent: Low
Ease of call back or follow-up: Difficult I Autormated )
 Speed of data collection: Instantaneous
3 , Onlime Speed of data collection: Instantaneous
Sample quality:
if selected randomly, quality better
• Cost effective if self-selected, quality poorer
A special type of self-administrated data collection • Geographic flexibility: Worldwide Geographic flexibility: Worldwide
E-Mail • Visual and interactive Cheaper distribution and processing costs
• Versatility of questioning: Extremely versatile Respondent cooperation: Varies if e-mail is seen as “spam”
Internet/ Web-based Surveys
• Questionnaire length: Individualized base on respondent answers
Mass mailing approach to contact responders. • Item non-response: Software can assure none Low Interviewers involvement
• Opt-in and opt-out option • Spam • Possibility for respondent misunderstanding: High • Mail, Internet (website/ email)
Pros:
• Speed, low cost • Interviewer influence of answers: None
• Supervision of interviewers not required • Eliminate the presence of interviewers
• Internet popularity for your population • Cheap, convenient, easier
• Anonymity of Respondent: Respondent can be anonymous or known
Richer survey design (video, random order, etc.) Cons:
• Special Features: Allows graphics and streaming media
Unable to response immediately (response quality could be lower)
Low response rate
,Se 1 f i mai7 panel dop off ]
4

1 -
Sample representative

Advantages
ObseroatonMethod Understand consumers without “asking” them.
Conditions for using Observation Alternative way to collect primary data
1. The needed information must be either observable or inferable from observable behavior. Disadvantages
2. The behavior of interest must be repetitive, frequent, or in some predictable manner. Public behavior only
• Otherwise, the cost of most forms of observation makes the approach prohibitively expensive. May not be projectable
3. The behavior of interest for many types of observation must be of relatively short duration. Time-consuming

Machime Observaton Hum am ObServation Benefits of Ethnographic research

→ Real world behavior:
Reality-based, show exactly how consumers live with a product, not just what they “say” or they “remember.”
Traf打o couutery Etbnographn

1 ⼁
_

It can show the members in a shopping trip, and know who is the decision maker.
Iu
- Storelfoof trafGr] Mystery shopper Explore the market:
May reveal unexpected needs and wants.
Neuromarketing Oue waymirvor
- May reveal some product problems.
Explore the competition:
Onlize
Can show how, when, why, and where people shop the brand among competitive brands.

ExDeriment Method → Cay sal velatiouship

It’s a research approach in which one variable is manipulated and the effect of another variable is observed or measured.
Strictly speaking, experimentation is NOT a form of data collection at all, but rather, a research strategy.
It is appropriate whenever we are attempting to determine the effect / impact of one thing on another.
7
叮
Man 5 pulaton r assign pp 1 9o differenf condi 9 ou ) establ 3 h canse rblatomship
.
Experiment vs. Survey and Observation
Research based solely on surveys or observations without any type of experimental design is strictly descriptive in nature.
(e.g., investigate “correlation” between two variables, or accessing people’s belief or agreement.)
Both survey and observation research, the researcher is a passive assembler of data. The researcher asks people or observes what they do.
In experimental research, in contrast, the researcher becomes an active participant in the process by “designing” an experiment.
Demonstrating Causation Y ( x+ Z ) + E
=

To demonstrate causation between variables X and Y (e.g., X causes Y), one must be able to show three things:
1. Correlation of concomitant variation.
• When X (does not) occurs, Y is (less) more likely to happen. They must vary together in some “predictable” way.
2. Appropriate time order of occurrence.
• X must happen earlier than Y
3. Elimination of other possible causal factors (e.g. extraneous variables).
• Even both (1) and (2) are satisfied, X may not cause Y.
• Another factor Z is causing X first, and Z also affects Y. In this case, X and Y has no causal relationship. Instead, Z -> X, and Z -> Y.
• So, we need to prove there is no “Z”. (Not feasible though!!)
The Key Variables in Experiment
#Dependent variables (DV)
Measures of effect or outcome that occur during the experiment, or measures of change in the conditions that exist after the experiment is completed.
Effect, outcome, result. (Y)
DV must be assessed after the IV is manipulated.
Usually relates to the responses from the consumers or markets (e.g., buying intention, time of spending, satisfaction, Like / Share/ Comment, etc.)
#Treatment variable / Independent variable (IV)
Variables whose value are directly manipulated by the researcher.
Cause, predictor. (X)
Typically, researchers design “two conditions” that vary in levels of the “IV”.
One is the experimental condition, and the other is a control condition.
The IV typically relates to 4Ps, or something that can be determined or controlled by the marketers. (e.g., price, package, music, etc.)

Experimental Validity refers to the manner in which variables that influence both the results of the research and the generalizability to the population at large.
Two kinds of validity:
• Internal validity: results of the research is valid and replicable.
• External validity: Generalizability of a study to a larger population, and outside of the lab.
lab experiments tend to have greater internal validity.
Research with high internal validity means that other researchers can replicate the experimental research easily.