This week I was happy to give a talk at the online conference of the International Society for Clinical Biostatistics about my work with Rachael Hughes on different ways of combining bootstrapping with multiple imputation for missing data. For those who may be interested, the video of this is now available on YouTube (below). For further details of our work, please see the published paper.
Online course - Statistical Analysis with Missing Data using R
I'm pleased to announce the release of a new (paid) online course - 'Statistical analysis with missing data using R'. In March I was due to go to Copenhagen to the Danish Cancer Society to give a short course on methods for handling missing data using R. COVID-19 put pay to the trip and the course and 2 months into lockdown in May I decided to try and turn the course into an online version. 3 months of night working later, and the course is complete!
I've used the Thinkific platform to host the course. It's a mixture of mini-lectures, multiple choice quizzes and video R sessions. The R sessions are embedded within multiple-choice quiz 'lessons', to enhance engagement with the content and to try and replicate, as far as possible, the in person experience. Each lesson has an attached forum so that participants can ask questions.
As well as what might be considered standard material on complete case analysis and multiple imputation, the final chapter considers more advanced multiple imputation techniques based to a large extent on areas I've worked on: in particular in regards accommodating the form of the substantive model when imputing missing values in covariates.
Bootstrapping multiple imputation using multiple cores/processors in R
I've written previously about combining bootstrapping with multiple imputation, in particular when the imputation and analysis models may not be congenial. This work has recently been published in Statistical Methods in Medical Research (open access paper here). The approach we recommend in this paper, proposed earlier by Paul von Hippel, is implemented in the R package bootImpute.
Read moreBootstrapping multiple imputation using multiple cores/processors in R