Red Door Analytics
Resources · Methods

By method

The same work, arranged the way a question arrives. Each page gathers the packages, the browser tools, the tutorials, the courses, our papers and the case studies for one method — instead of leaving you to find them across four sections.

The core method

Survival analysis

Analysing the time until something happens, when for some people it has not happened yet. Those people are censored rather than missing — they carry information right up to the moment they leave, and the methods here are the ones that use it.

12 tutorials · 2 courses · 77 papers · 1 case study · 2 packages · 3 tools

Open survival analysis →
Narrower methodsCompeting risks7Flexible parametric survival models2Multi-state models5Recurrent events3Relative survival1
Method

Joint longitudinal–survival models

A repeatedly measured biomarker and a time-to-event outcome fitted as one model, because the trajectory is what drives the risk. Measurement error is carried rather than ignored, and follow-up stops informing the trajectory when the event happens.

5 tutorials · 2 courses · 17 papers · 1 case study · 1 package

Open →Method

Longitudinal data analysis

Repeated measurements on the same people, where observations within a person are not independent. A mixed model separates the population trend from each person's departure from it, which is what keeps the interval around that trend honest.

2 tutorials · 3 courses · 19 papers · 1 case study · 1 package

Open →Method

Simulation studies

Generating data from a truth you chose, so a method can be judged rather than trusted. It is how bias, coverage and power are measured, and how a study is designed before anyone is recruited.

5 tutorials · 19 papers · 2 packages · 1 tool

Open →Method

Causal inference

Estimating what would happen under one intervention rather than another, from data where people were not randomised. The estimand is a contrast between two versions of the same cohort, and it rests on assumptions the model cannot check.

2 tutorials · 2 papers · 1 case study

Open →Method

Prediction and risk modelling

Building and validating a model that gives an individual their risk. Discrimination says whether the ordering is right; calibration says whether the numbers are, and a model can get the first right while getting the second badly wrong.

2 tutorials · 9 papers · 1 case study · 3 packages · 1 tool

Open →Method

Health economic modelling

Turning survival into the life-years and quality-adjusted life-years a decision is made on, usually far beyond the trial's follow-up. The extrapolation is the part that moves the answer, so it is the part that has to be defensible.

2 tutorials · 5 papers · 1 package · 2 tools

Open →
Not yet

In the vocabulary, not yet on the site

These are methods we work in and have not published on. They are listed because a gap you can see is worth more than a tidy page.

  • Missing dataHandling values that were not recorded, in a way that says what is being assumed about why. Complete-case analysis makes an assumption too — it just does not state it.
  • Register-based researchAnalysis built on national registers, where the cohort is a whole population and follow-up is near-complete. The strengths and the traps both come from data collected for another purpose entirely.
  • ReproducibilityMaking an analysis something another person can run and get the same answer from. It is a property of the code and the data trail, not of the intentions of whoever wrote it.