Posts – Notes

The c-index and AUC-ROC for competing risks outcomes

Coding the metrics in R

In a previous post I looked at how to generalise the concordance index (c-index) and area under the receiver operating characteristic curve (AUC-ROC), to competing risks…

Oisin Fitzgerald

Evaluating discriminatory ability of prediction models for competing risks outcomes

Background and theory

A note on scope: this post focuses on discrimination, the ability of a model to rank subjects correctly. A model can have excellent discrimination while being miscalibrated…

Oisin Fitzgerald

Basics of Bayesian modelling

With R code examples

Bayesian inference is a method of statistical inference in which we use probability to expresses a degree of belief or uncertainty about the parameters in our statistical…

Oisin Fitzgerald

Missing data and directed acyclic graphs

Notes on Mohan & Pearl (2021)

In many data science applications it is quite common for there to be a need to account for missing data. Missingness graphs (or m-graphs) extend traditional causal graphs by…

Oisin Fitzgerald

Binary outcome misclassification

Nondifferential misclassification

Misclassification of binary outcome data can lead to biased estimates of model coefficients when using logistic regression. In certain situations where we have information…

Oisin Fitzgerald

An introduction to forward mode automatic differentiation

With example Julia code

A look at the first half (up to section 3.1) of the survey paper by Baydin et al. (2018)

Oisin Fitzgerald

Causal Mediation: A Brief Overview

With some R coded simulations

I recently worked on the meta-analysis of a set of randomised control trials (RCT), and as part of this carried out a series of causal mediation analyses. I’d never thought…

Oisin Fitzgerald

Bitcoin price volatility with ARCH models

This post covers some of the basic strategies behind how (financial) time series are analysed and how volatility models work. In particular I examine the ARCH model. Don’t…

Oisin Fitzgerald