Curriculum Development

Introductory Multivariate Analysis

In 2023 the department introduced Introductory Multivariate Analysis (STA739), a renaming of STA734 to emphasise the continuation of the students' multivariate analysis journey in the second semester with STA701 (Multivariate Analysis). Most statistical analyses form part of multivariate analysis and STA739, together with STA701, form our students' multivariate analysis journey. Given my extensive experience with teaching and supervising in the Data Science Masters programme at this time, I was in a position to review all topics covered by our other Honours courses and identify topics that are missing, but relevant.  Relevant here meaning to the knowledge set the students should have going from Honours into the Masters programme, but also techniques currently applied by career analysts. Thus, considering my students' learning needs and making sure they feel prepared when moving from my Honours programme to the next stage of their analytics journey. This content evaluation led to the following topics being covered in STA739 in the first semester:

1. Multivariate numerical data analysis

2. Multivariate categorical data analysis

3. Survival analysis

The rest of the topics are covered in the second semester Multivariate Analysis course (next section). These topics were selected for the first semester course since my research of relevant literature suggested that a good foundation on these concepts are necessary before moving on to the other topics.

Traditional multivariate analysis textbooks are focused on continuous data. In the real world, Statisticians and Data Scientists work with mixed data types (numerical and categorical). Also, survival analysis is a separate and extensive set of tools for the analysis of time-to-event data and as such will not be included by typical multivariate analysis textbooks. As such, after identifying these topics, I authored course readers for each topic from the following main sources:

• Johnson, R. A., & Wichern, D. W. (2007). Applied Multivariate Statistical Analysis (6th ed.). Upper Saddle River, New Jersey: Pearson Prentice Hall

• Rencher, A. C., & Christensen, W. F. (2012). Methods of Multivariate Analysis (3rd ed.). Hoboken, New Jersey: John Wiley & Sons

• Agresti, A. (2013). Categorical Data Analysis (3rd ed.). Hoboken, New Jersey: John Wiley & Sons, Inc.

• James, G., Witten, D., Hastie, T. & Tibshirani, R. (2021). An Introduction to Statistical Learning with Applications in R.

I also reviewed hundreds of peer-reviewed articles to identify top quality relevant foundational content. Those that I selected, along with the main sources, can be viewed in the bibliographies of the course readers:

1. Multivariate numerical data analysis sources

2. Multivariate categorical data analysis sources

3. Survival analysis sources

Looking at the contents of the course readers over the years (2023 - 2025), it is noticed that I've been reviewing and updating the content annually. The reason for doing this is two-fold. Firstly, I do this to ensure that the topics remain relevant. Updates in this regard are informed by research and supervision that I do. For example, point-biserial correlation and Cramer's V were added to the 2025 categorical data analysis course reader (↪ view) following the industry-directed project I supervised in 2024, where effect size measurement was required in addition to knowing whether significant associations exist. Also, I make these updates to ensure that the content is as clear and easily digestible as possible for the students. 

The structure of these course readers is an introduction to a concept, followed by mathematical formulation and/or derivation, and then practical application, first by hand to cement the understanding of the technique, and then in RStudio (↪ view). Part of the reason for compiling my own course readers, is to add software application of the techniques throughout. Traditional textbooks do not always follow this structure that I am after. Either they focus on theoretical and mathematical discussion with hand calculations, or they provide summary high-level discussions of techniques followed by software application. With the structure of my course readers my students get to know a new topic before going into the mathematics behind the technique. Then they see how the mathematics are applied, by hand. And, finally, how to apply it in a software package. The course readers also include links to the helpfiles of the functions used to apply the analyses. These are the words highlighted in blue (↪ view). I follow this approach in all course readers that I author. A student commented in 2024: "I enjoyed reading all the course materials because they were very self explanatory".

Statistics graduates are now expected to develop a relevant technology stack. As such my curriculum development needs to take this into account to ensure I deliver analytics graduates with the desired attributes. In Statistics and Data Science, this would be SAS, Python, and RStudio. Our postgraduate students receive training in Python and SAS in other modules. I train them in using RStudio. This training goes beyond coding in base R. Students are shown how to use scripts with comments to contain all analyses relevant to a particular task, in one place (↪ view). Students now also use R Markdown to compile their assignments (↪ view). As explained by by Wickham et al. (2016), R Markdown "provides an unified authoring framework, combining your code, its results, and your prose commentary". The analyst then has the option to knit the Markdown file to a wide selection of output formats, such as Word, pdf, Powerpoint, etc. R Markdown also makes it possible to switch between engines (R, Python, SAS) in the same file, which is something I also make them aware of. RStudio can thus be used as a general analytics workbench (↪ Python, ↪ SAS). My curriculum development for this module and the follow-on Multivariate Analysis module (next section) takes this into account. 

Further curriculum development includes the integration of GitHub Copilot for RStudio in 2026. This technology has the potential to accelerate code development and provide students with a practical understanding of how AI tools are being utilised in programming environments, another essential skill for their growing technology stack. This would require an update to how assignments are designed to assess their practical skills development (↪ view).

Of course, in addition to the course readers I also prepare slide decks for lecturing (↪ view) and R scripts for demonstration. Since, at this stage, the students aren't necessarily familiar with Python, I only focus on R. However, in Multivariate Analysis (next section) I have gradually added Python application as well based on my own skills development using this package.

Multivariate Analysis

This second semester Multivariate Analysis module follows on from the Introductory Multivariate Analysis (previous section). The following topics are covered in this course from the topics I identified:

• Overview of matrix algebra required for multivariate analysis

• Principal component analysis (PCA)

• Principal component regression (PCR)

• Discriminant analysis (descriptive/exploratory)

• Classification analysis (linear and quadratic discriminant analysis)

• Hierarchical cluster analysis

• K-means cluster analysis

In addition to designing the course content, it is my responsibility to demonstrate the application of these techniques in RStudio. When I am given this task, my focus is on turning my students into analytical problem solvers. I make sure that they understand the process they need to follow from raw data to results through software application. With data-driven decision making at the fore, it is also important that my students understand how to turn the results into insights relevant to the problem being solved. In order to achieve the desired outcomes, I thus also authored the following practical course readers for this purpose:

• Matrix Algebra Practical

• PCA and PCR Practical

• Discriminant Analysis Practical

• Classification Analysis Practical

• Cluster Analysis Practical

Lists of the sources consulted in the compilation of these course readers are included in the texts. The matrix algebra course reader also contains Python code in addition to the R code. This is currently being done for the other remaining course readers.

Here too I incorporate ideas from research and supervision. For example, the PCA and PCR manual includes a section on what students need to do if they have mixed type data, given that these techniques are traditionally applicable to numerical data. This is not simply given as self study, I make a point of thinking through this with them during the lecture (↪ view) since this is likely an issue they will be confronted with in practice. 

With cluster analysis, traditional textbooks stop at demonstrating how the technique is applied. When my students enter industry, they need to do more than that. Cluster analysis solutions are based on a series of subjective decisions. My students understand this and know that they will typically end up with multiple possible solutions from which the "best" solution must be selected. And then they will need to turn the "best" solution into actionable insights for their client. So, in addition to demonstrating the cluster analysis techniques, I also include cluster validation and interpretation (↪ view). 

During practical demonstrations the students are given opportunity to practice the R application (see Hands-on Practice for an example) in class and receive short assignments to practice on their own afterwards (example). Towards the end of the semester the students receive a practical project bringing all the topics together to test their understanding of where each technique fits in and what insights can be deduced by applying them.

When my students pass Multivariate Analysis, I know that I have given them every possible resource they need to be able to apply what I've taught them in a practical and insightful way.

Time Series Analysis

After the first cohort of MSc Statistics (with specialisation in Data Science) students graduated, the Data Science lecturing team reflected on the MSc curriculum that was taught and industry projects that were completed.  It was decided to revive the Honours level Time Series Analysis (COF711) module given the wide application of this methodology to problems in the financial and retail industries.  The task was given to me to redevelop the time series analysis curriculum, given that the module had been inactive for a while. After some consultation with peers that had or were presently lecturing the topic, I identified an appropriate textbook for the course. It came highly recommended as it was free, interleaved software application in RStudio with theory and covered a range of topics appropriate for a first level postgraduate course on the topic. 

In a nutshell, time series forecasting using exponential smoothing methods and ARIMA models formed the basis of this module.  I covered Chapters 1 - 4 and 7 with the students and, in addition to the R application included in the textbook, I also incorporated equivalent SAS application. For an example of a lecture, ↪ view.

Statistics and Visualisation

In 2017, developed for the EMS Faculty's Diploma in Software and Media with specialisation in Data Analytics and Business Intelligence, the department started offering BIA713, Statistics and Visualization.  At this time I was asked to develop the curriculum for this course. The student cohort typically consisted of a mixed set of professionals and students with varying levels of Statistical knowledge and computer skills that needed to be taken into account. Also, the objective was for the taught content to be applicable for them in the environments they work in. I curated the topics for this first course in data analysis keeping these points in mind. I decided on the following topics (↪ view):

• Introduction to data

• Probability (self-study - perhaps too advanced)

• Distributions of random variables

• Foundations for inference

• Inference for numerical data

• Inference for categorical data

• Introduction to linear regression

• Logistic regression

I developed all course material for the parts that I lectured (everything up to logistic regression), which included interleaving computer application with theory content. 

An crucial aspect of the course was that the students must learn how to gain insights from the data analysis we teach them.  For an example of my lecture notes, ↪ view.  In addition to this, it was also important that they learn to apply the data analysis in software packages that would be readily available to them where they work. As such, I decided to focus on Excel, as most businesses at a minimum have access to Excel. Then I also decided to show them how to do data analysis in R, since it is open source software that can be downloaded and installed for free. Due to the dual software application and wanting to help them upskill as seamlessly as possible, I compiled a manual with clear instructions on how to carry out the various statistical techniques in Excel and/or R (↪ view).

I thoroughly enjoyed presenting this course, even though it could be challenging navigating different generations with varying levels of analytics ability in the same class. The appreciation though for the effort made, was priceless. Below a photo of me presenting one of my lectures in Lab J of the CAMS building in 2018.