Ilah Waals & Larissa van ‘t Westende
Lecture 1: Introduction
Marketing research (process): the systematic and objective identification, collection, analysis,
and dissemination of information for improving decision-making related to the identification and
solution of problems and opportunities in marketing.
Problem definition:
- Decision problem (focus on action): How should we position our product on supermarket
shelves to attract maximum consumer attention?
- Research problem (focus on understanding): How does shelf positioning affect consumer
attention.
Research approach:
- Exploratory research goal: may be rule-based or data-driven.
- Explanatory research goal: a conceptual goal should be guided by theory. In simple
terms, a research model determines the relationships between different variables.
1. Main effect: Does the IV influence the DV?
Hypothesis must clearly include (a) all variables and (b) direction of the relationship.
2. Moderation effect: The direction or strength of the effect of IV on DV is affected by a
moderating variable (or moderator)?
3. Mediation effect: Whereas moderator variables specify when certain effects will hold,
mediators speak to how or why such effects occur.
Research design: Determines the information that’s needed to answer the specific research
questions, or to test the developed conceptual model and hypotheses.
- Takes into account:
- Nature of the issue: common behavior (spontaneous response); personal or
sensitive issue, repressed tendency
, - Nature of the respondents: age, background, previous participation
- Context: cultural norms, ease of data collection
Data collection:
- Secondary Data: (collected for some other purpose than problem at hand):
- External: governmental, non-governmental data
- Internal: customer data
- Primary Data: (collected primarily for purposes of problem at hand):
- Quantitative methods: surveys, panels, descriptive data
- Qualitative methods: in-depth interviews, focus groups, ethnography, observation
- Causal research methods: lab and field experiments
Data analysis:
- Quantitative methods: basic analyses, analysis of variance and covariance, survey
techniques, item analysis, factor analysis, regression, cluster analysis, multidimensional
scaling, conjoint analysis
- Qualitative methods: content analysis, semiotics
Report & present:
- Report the results of each stage of the marketing research process
Vector: the simplest type of data structure in R. A single entity consisting of a collection of
things. Members of a vector are called components.
- Example: c (“Amsterdam”, “Rotterdam”, “Utrecht”); c (4, 5, 9, 3, 6)
Factor variable: You can make categorical variables into factors in order to be able to use them in
R.
Descriptive Analysis: type of analysis of data that summarizes variables of a dataset.
- Example: means or bar plots
, Lecture 2: Basic Data Analysis
*likert scales
Notations of statistical hypotheses: Use H𝑛𝑢𝑙𝑙 and Halt to avoid confusion with theoretical
hypotheses H0 = null and H1 = alternative
Hypothesis Testing: Two-sided tests
- H𝑛𝑢𝑙𝑙 : the parameter (e.g., mean, proportion) of the variable is equal:
- One variable or univariate (e.g., age): mean age = 40 (in this sample, the average
age is equal to 40)
- Two variables or bivariate (e.g., age or gender): mean agewomen = mean agemen (in
this sample, the average age for women and men are equal)
- Halt : the parameter of the variable is different:
- One variable or univariate (e.g., age): age ≠ 40 (in this sample, the average age is
different than 40)
- Two variables or bivariate (e.g. age and gender): mean agewomen ≠ mean agemen (in
this sample, the average age for women and men are not equal)
Hypothesis Testing: One-sided tests:
- Hnull: the parameter of the variable is > or < : mean age > 40
- Halt: the parameter of the variable < or > : mean age < 40
Hypothesis Testing: p-value
All statistical tests result in a test statistic & p-value (significance level):
- The lower the p-value, the greater the statistical significance of the observed difference.
- If p < 0.5, H𝑛𝑢𝑙𝑙 is rejected → the parameter is significantly different from a
specific value (univariate test) or there is a significant relationship between two
variables (bivariate)
, - If 0.5 < p < 0.10, H𝑛𝑢𝑙𝑙 is rejected but marginally → e.g. there is a marginally
significantly relationship between two variables (bivariate)
- If p > .10, H𝑛𝑢𝑙𝑙 is not rejected → the parameter is not statistically different from a
specific value or across groups
Hypothesis testing: other decision criteria
Test statistic (inversely related to p-value): When the absolute value of the test statistic > critical
value, the null hypothesis is rejected. Critical values are obtained in statistical tables.
- For example, in a t-test, if ltl > 1.96, the null is rejected: there is a significant difference
between groups
- For example, in a t-test, if ltl < 1.96, the null is not rejected: there is no significant
difference between groups
95% confidence interval: When the null hypothesis posits that a parameter estimate (f.e.
correlation coefficient, regression coefficient, mean difference) is equal to 0 (or another number):
if the 95% CI of that parameter estimate excludes 0 (or the other number), then we reject the null
hypothesis.
- If the correlation coefficient is .19 with 95% CI (.03 to .35): the null is rejected
- If the correlation coefficient is .17 with 95% CI (-.02 to .30): the null is not rejected
Statistical Analyses: Univariate test (one variable)
- Example RQ: Testing whether the sample is representative of the general population for a
given variable
- Often, these tests are used to examine whether the sample is representative of the general
population
