SKI3011 syllabus

Evidence Synthesis 2: Statistics in Systematic Reviewing

This module

In SKI3011 you choose your own research topic and carry out a systematic review and meta-analysis from start to finish. You write a protocol and register it on protocols.io, manage screening and data extraction in Covidence, and learn to run meta-analyses in R: effect sizes for binary and continuous outcomes, proportions and correlations, forest plots, heterogeneity, publication bias, subgroup analysis and meta-regression. The pace is higher than in SKI3010, and you get more freedom.

The first six weeks are carbon-based. You write your own R code, and each week a short paper quiz checks whether you can read and interpret the code and output from the previous lecture. In week 7 the module turns silicon-based with a demonstration of Claude Code. From then on you may also let AI do part of the work in your own project, but you do not have to. You finish with a draft paper (introduction, methods, results) plus R code and dataset, which you complete in PRO3017.

Learning outcomes

After this module you can:

  1. formulate a review question and register a review protocol;
  2. manage screening and data extraction in Covidence;
  3. read, write and interpret R code for meta-analysis of binary, continuous, proportion and correlation data;
  4. assess heterogeneity and publication bias, and explore heterogeneity with subgroup analysis and meta-regression;
  5. write a draft systematic review with reproducible R code and dataset;
  6. give constructive feedback on a review and its code written by others;
  7. explain how an AI tool such as Claude Code can support a meta-analysis, and check its output.

Assessment

Component Weight Notes
Quizzes 1-5 (paper, in the tutorial, interpreting R code and output) 25% Best 4 of 5 count
A4 Individual peer review 15% Individual
A5 Final paper + R code + dataset 60% Group of 1-3 students
A1-A3 Progress assignments 0% Pass/fail

The final paper and the peer review are graded with the rubrics.

Rules

  • Groups. You work in a group of 1 to 3 students on a topic of your own choice. Every member is responsible for the whole paper and code. The peer review is individual.
  • Quizzes. Paper quiz at the start of the tutorial, no devices. The best 4 of 5 quizzes count. A missed quiz scores 0 and has no separate resit. With documented illness you can take the quiz in the next tutorial.
  • Deadlines. Hand in via the submission form. Always include your R script and your dataset (csv), so that your results can be reproduced.
  • AI is allowed as a tutor from week 1 and as a co-worker from week 7; see AI use. Every paper contains a short AI-use paragraph in the methods.
  • Announcements are sent through Canvas and reach you by e-mail.