The module will consist of two sections with each addressing the need for some familiarity and competence in available and useful techniques in two data analysis contexts, namely; survey analysis & predictive modelling.
The first section shall comprise of standard statistical techniques useful in analyzing typical survey type data; which can consist of different measurement levels. This will be focused towards conferring on the student a useful repertoire of statistical techniques for analyzing data generated by their own surveys which shall form part of a group project which addresses a research question in a context of their own choosing. this will also support thesis data analysis in year 4 dissertations.
The second section shall look at some techniques that are useful in predictive modelling and in particular the prediction of a categorical target (a classification problem) using techniques such as logistic regression & classification trees as well as regression for numerical targets.
Classes will be lab based and therefore facilitate a hands-on approach with sample data available for analyses and will involve primarily the usage of SPSS.
The group project shall consist of the students identifying a topic of their choosing and subject to agreement to then develop research objectives, design and deliver a survey and subsequently analyze the data and generate insights, conclusions and recommendations via a formal report. This process will be supported through class mentoring, provision of a timeline and feedback at various stages.
Module aim
It is intended that the student will be capable of identifying some research question and subsequently to construct a fit for purpose survey to address the research objectives and to subsequently identify, plan and conduct statistical analysis pertinent to the research objectives and the type of data generated.
It will be the aim to develop this statistical analysis capacity through a range of practical data contexts and through lab-based usage of SPSS and a range of sample data files.
A further aim is to facilitate some self-directed learning by facilitating a project characterized by self-selecting groups, self-selecting topics of interest to the group, a timeline with mentoring and feedback but ultimately an ownership of the process by the student where they choose the survey tools and the statistical techniques and produce a formal report.
A further aim is to introduce and engage the student in more modern predictive analytics and modelling approaches and opportunities and to appreciate the ever emerging data- rich environments marketers find themselves in and the need for sophisticated analyses for insight generation and decision support
The module shall firstly address some of the more traditional statistical techniques pertaining to survey analysis with some bivariate and multi-variate techniques and hypothesis testing reviewed and extended from previous years. It will also look at some predictive modelling & classification approaches that arise in more modern data-rich customer environments and CRM contexts. This material will be developed through a range of data examples and will have an applied and practical emphasis with marketing implications discussed.
Introduction: (self directed reading elements) the volume, variety and velocity of data in modern marketing. The need and occasions for analysis in marketing. Analytical approaches and tools, metrics, dashboards, big data.
Data types: a review of different measurement levels, simple SPSS features
Simple analysis options: Custom tables, pivot tables, OLAP cubes, data explore, graphs etc.
Association Analysis: Review of tests of association, 2x2 tables, larger tables- test for independence-Pearson measures of association, measures of strength (phi, Cramer’s etc.) comparing proportions, odds ratios, relative risk etc. Application to marketing and survey data.
Comparing means: assumptions of Anova, tests of normality, homogeneity of variance test, standard F-test of Anova, follow up tests. Robust Anova- Browne –Forsyth & Welch test and follow up tests-Games-Howell.
Non-parametric tests: when to use them, assumptions. Kruskal-Wallis test & median test, follow up tests
- Time permitting possibly also Friedman test, Wilcoxon test, Cochran & McNemar tests.
Predictive Analytics & scoring models: classification approaches- tree based approaches such as CHAID, CRT etc, logistic regression and neural networks – data examples and analyses – confusion matrices and performance metrics such as sensitivity, specificity, gains charts AUC etc. comparing models and methods. Scoring, segmentation, profiling and targeting
Software tools such as SPSS, BigMl machine learning, marketing applications and examples.
Multivariate methods & Survey Analysis:
Cluster Analysis: Segmentation and profiling, k-means clustering, hierarchical clustering, linkage methods, profiling and labelling clusters. Contrasting clusters and hypothesis tests.
Case study and survey applications. Marketing implications.
| A recognition of the potential concern sometimes articulated by students regarding statistical analysis will translate to the adoption of a measured progression through a range of simple examples and reviews with the encouragement of questions in an informal setting and encouragement to conduct the analyses themselves both in class and later as suggested homework tasks. The approach will be one of a lab-based setting; allowing for the usage of software tools such as SPSS etc Students will have access to adequate sample data sets and tutorial material relevant to the types of techniques used and will be tasked with modest homework tasks to consolidate class delivered material Significant learning will also be developed through the group project process and this will be supported through a timeline approach with deliverables and deadlines in a mentored process where feedback is given and time allocated to problem resolution at key points such as survey drafting, spreadsheet preparation, analysis planning and implementation. Some past projects will be provided as a resource and to encourage improved standards also accelerate the learning curve and provide a benchmark and driver for quality. An element of self-directed study will also be a feature of the course with directed/suggested usage of web in particular around predictive modelling Where possible it is further envisaged that certain journal articles that utilize the techniques in an accessible way to students will be provided for reference and good practice examples Attendance at classes is expected and there shall be an attendance requirement with unexplained absences resulting in a loss of marks- typically 10% if exceeding a threshold- to be advised in class.
|
| Module Content & Assessment | |
|---|---|
| Assessment Breakdown | % |
| Formal Examination | 60 |
| Other Assessment(s) | 40 |