Midterm 1 Flashcards
What is Marketing Research?
A collection of techniques for collecting, recording, analyzing and interpreting date for business decision making.
Backward marketing research
Research done from the end, product-forward. Produces action-oriented results.
How to qualify marketing research
Reliability, cost, potential profits with/without info, management’s ability to use info
Expected profit
E(profit) = probability*profit1 + probability*profit2 E(info) = E(Profit + Info) - E(Profit)
Exploratory Research
Used when there is an ambiguous problem. Requires secondary data, qualitative data.
Descriptive Research
Used when the problem is somewhat defined, need to know how to solve it. Requires a survey.
Causal Research
Used when problem is clearly defined to find solutions. Requires an experiment.
Primary data sources
Benefits: Up to date, perfect fit for questions, control over how it is collected.
Qualitative: Focus group, interview
Quantitative: Survey, Experiment
Secondary data sources
Benefits: Cheaper than primary data, readily available. Good starting point.
Sources of Data
Scanner Data: literally use scanners, see who’s buying what. Can’t see why
Single-source Data: Nielsen ex. Use stream of data from single consumer (panel)
Geo-segmentation: Segmentation by demographics, lifestyle
Attitudes, Intentions, Behavior
Attitude: a positive or negative evaluation of a product
Intention: an indication of an individual’s readiness to perform a given behavior
Behavior: an individual’s observable response
When to use qualitative research?
Exploratory studies: to establish basis for quantitative research
New product development: after quantitative research to identify gaps
Focus groups: Pros and cons
Pros: Easy, good for in-depth info, complex issues can be discussed, one person’s experiences stimulate others
Cons: Superficial reactions, not quantifiable, group process may stunt frank exchange, minority viewpoints may not be heard
Zaltman Metaphor Elicitation Technique
Uses metaphor to reveal unconscious thoughts and thoughts based on sensory info
Primary Data: Qual vs. Quant
Qual: Small # of cases, unstructured data, nonstatistical inference, give richer understanding
Quant: large # of cases, structured, statistical inference, recommends final course of action
Types of scales
Nominal: Categorical. Measures frequency of recorded variable. Ex: Male or female
Ordinal: Categorical. Can be ranked. Ex: 1-5 min, 6-10 min, etc.
Interval: Quantitative. Data can be treated as numbers. Ex: strongly disagree - strongly agree (1-5)
Ratio: Quantitative. Fill-in the blank, numbers are numbers. Ex: what is your annual salary? ___
Central tendency
Mean.
Dispersion
Variance/standard deviation
Ratio scales
Likert: agree - disagree
Rank-order: rank items by importance
Paired comparison: which of these two do you prefer?
Requirements of causal inference
- Correlation
- Temporal Antecedence (x must come before y)
- No third factor driving both
What kind of data can be used to determine causation?
Only experimental data. Correlational data (scanner, surveys) cannot be used.
Reliability
Will I get the same result if I measure again? Affected by random error.
Validity
Am I measuring what I am supposed to measure? Affected by systematic error.
Observed v. True Score
O = T + e(systematic) + e(random)
Internal validity
Accuracy, measuring what you meant to. Extent to which results reflect the truth.
External validity
Can you generalize the results? Extent to which results will hold beyond experimental setting
Threats to internal validity
History effect Maturation effect Pre-test effect Instrument Variation Statistical regression Selection effect Mortality
Threats to external validity
History effect Maturation effect Pre-test effect Instrument Variation Statistical regression Selection effect Mortality Reactive bias Non-representative sample, environment, materials used
Between-subjects design
Experimental design in which each subject only receives one treatment. Comparisons are made between different groups of subjects
Within-subjects design
Experimental design in which subjects receive more than one treatment. Statistically superior, but not always possible.
Experimental design notation
O: Any observation or measurement X: Exposure of experimental units to the treatment EG: Experimental group CG: Control group R: Random assignment
Interaction Effect
Effect of one independent variable on the dependent variable changes depending on the level of another independent variable
Main Effect
The effect of one of the independent variables averaging over all levels of other variables
Types of Sampling
Non-probability sampling: Population elements are sampled in a non-random manner. Exploratory stage, pre-test, cost effective
Probability sampling: Every element has a known, non-zero probability of inclusion in the sample
Non-probability sampling
Judgmental sampling: use experts’ judgment to identify samples
Snowball sampling: One respondent identifies other respondents (small, specialized community)
Quota sampling: sample a minimum number from each specified subgroup in the population
Probability Sampling
Simple random: Everyone has the same known nonzero probability of inclusion
Stratified: Population is divided into strata. Sample a proportion of each strata (either equal or based on size or variability)
Cluster: Population is divided into clusters. Each cluster is selected randomly. Everyone within cluster is a respondent.
Target Population
Define the relevant population
Sampling Frame
List of elements from which the sample may be drawn
Sampling Unit
Group that is selected for the sample
Mean (Mu, x bar)
Population mean = (1/N)(X1 + X2+…+Xi)
Sample mean = (1/n)(X1 + X2+…+Xi)
Variance (omega^2, s^2)
Pop = (1/N)[(X1-popmean)^2 +…+(Xi-popmean)^2]
= SD^2
Sample = (1/n-1)[(X1-sampmean)^2 +…+(Xi-sampmean)^2]
Standard deviation (omega, SD)
SD = square root of variance
Normal distribution
1 SD from mean: 68%
2 SD from mean: 95%
3 SD from mean: 99.7%
Standard deviation of the sample mean
SD/(square root of sample n)
Confidence interval
68% confidence interval: (Xbar -SD, Xbar +SD)
95% confidence interval: (Xbar -2SD, Xbar +2SD)
99.7% confidence interval: (Xbar -3SD, Xbar +3SD)
OR
(Xbar -ZSD, Xbar +ZSD)
Z = 1, 1.96, 2.58