Marketing Research

1. General Marketing Research Process

3. Problem Formulation

5. Data Collection

7. Measurement

9. Sample Size

11. Multi Regression Analysis

13. FACTOR Analysis

2. Total error

4. Research Design

6. Data Collection Form

8. Sampling

10. Data Examination

12. ANOVA

14. Structured Equation Model

1. General Marketing Research Process Back to Index

1) Formulate marketing research problem
2) Determine research design (Exploratory / Causal)
3) Determine data collection method (Secondary / Primary)
4) Design data collection forms
5) Design sample and collect data

Sampling frame
Sample selection process (Probability sample / Nonprobability sample)
Sample size

6) Analyze and interpret the data

Editing (Check the completeness and consistence of data as well as completeness of instructions)
Coding (for computing)
tabulating

7) Prepare the research report

2. Total Error Back to Index

Larger sample does not necessary increase accuracy. Larger sample may increase total error.

3. Problem Formulation Back to Index

1) Translate Decision Problem into Research Problem
2) Research Proposal

1.Tentative project title
2.Statement of the marketing problem
3.Purpose and limits of the project
4.Outline (tentative framework of the project)
5.Data sources and research methodology
6.Estimate of time and personnel requirements
7.Cost estimates

4. Research Design Back to Index

Exploratory Research (General an starting research)

Literature search
Experience survey
Focus group
Analysis of selected cases

Descriptive Research (Relationship finding research)

Longitudinal (Continue to pick up the same categories)

True Panel (Continue to pick up the same persons, ex. Brand switch rate)
Omnibus Panel (Pick up different persons at each time)

Cross Sectional (Pick up different categories & different persons at any time)

Causal Research (Cause & effect finding research)

Data Collection Back to Index

1) Secondary data

2) Primary data

a. Communication

Degree of Structure (Structure or unstructured)
Degree of Disguise (Data source knows purpose of survey or not)
Method of Administration (Interview, telephone, mail questionnaire)

b. Observation

Degree of Structure (Structure or unstructured)
Degree of Disguise (Data source knows purpose of survey or not)
Setting (Natural or Contrived (=artificial situation))
Method of Administration (Human or mechanical)

6. Data Collection Form Back to Index

7. Measurement Back to Index

Scales

	Scale	Examples
	Nominal	Male/Female User/nonuser Occupations
	Ordinal	Preference of brands Social class Graded quality of lumbers
	Interval	Temperature scale
	Ratio	Units sold Number of purchasers Probability of purchase Weight

Eliminate errors as much as we can.

1) Classification of Errors

Xt=TRUE SCORE

Xo=OBSERVED SCORE

Es=SYSTEMATIC ERROR (a constant error, ex. measure is not accurate)

Er=RANDOM ERROR (a transient error, ex. shoes on or off when we measure height)

Es+Er=TOTAL ERROR

Xt=Xo+Es+Er

When a measure is VALID, Es+Er=0 ® Xt=Xo

When a measure is RELIABLE, Er=0 ® Xt=Xo+Es

2) Assessment of Validity

Content validity: The adequacy with which the domain of the characteristic is captured by the measure.

Construct validity: Assessment of how well the instrument captures the construct, concept or trait it is supposed to be measuring.

Convergent validity: The confirmation of relationship by independent measurement procedures

® Independence of each measurement procedure

Discriminant validity: Requirement that the measure of construct does not correlate too highly with another measures from which it is supposed to differ

® Independence of each construct

3) Assessment of Reliability

Stability: Small difference between two different time points of the identical construct.

Equivalence: Adequate correlation among all items answered by the one person.

Coefficient a : Summary of intecorrelations among a set of items.

k=# of items s _I=variance s _t=total variance

8. Sampling Back to Index

Sample drawing procedure

Define the population
Identify the sampling frame
Select a sampling procedure
Determine the sample size
Select the sample elements
Collect the data from the designated elements

Sampling procedure

Probability samples

Simple random
Stratified

Proportionate
Disproportionate: Homogeneous population needs small samples to get statistic estimates.

Cluster

Systematic sample: choose every k th element
Area sample: Pick up reasonable group as a convenient population. (ex. college for college seniors, airports for airline travelers)

Non probability samples

Convenience (accidental) samples
Judgement samples: select the sample which is believed to represent a population.
Quota: use a proportionate element with a parent population to draw samples.

Sample distribution

68.26 percent of the sample means will be within +-1 s _x‾ of the population.

95.45 percent of the sample means will be within +-2 s _x‾ of the population.

99.73 percent of the sample means will be within +-3 s _x‾ of the population.

9. Sample Size Back to Index

When a population variance is unknown:

The half of the interval inference = z s _x‾

When population is probability,

10 Data examination Back to Index

Outliers

Normality

Heteroscedacity

Skewness

Linearity

11. Multiple regression Back to Index

R square: coefficient of determination, how much the variation from the regression can be explained by X.

Beta coefficient: relative impact of each coefficient (coefficient for each standard error)

Variables ® a multiple variate

Assumptions:

Linearity

Normality

ANOVA Back to Index

Assess whether the difference between group means is significant or not

T-test:

Assumptions:

Dependents must be independent among each other.

Normality

Equality of covariance matrices

FACTOR Analysis Back to Index

Explanatory research

Find factors which highly correlate to variables

14. Structured Equation model Back to Index