Kylie Lange
@klange-au.bsky.social
Statistician working in clinical trials methodology & nutrition and diabetes clinical research | South Australian Health & Medical Research Institute (SAHMRI); Adelaide University
This is work from The University of Adelaide, @sahmri.bsky.social and @austrim-cre.bsky.social, with big credit and thanks to my co-authors and supervisors, Lisa Yelland, Tom Sullivan and @jesskasza.bsky.social
February 5, 2026 at 6:43 AM
This is work from The University of Adelaide, @sahmri.bsky.social and @austrim-cre.bsky.social, with big credit and thanks to my co-authors and supervisors, Lisa Yelland, Tom Sullivan and @jesskasza.bsky.social
We know that many partially clustered trials do not account for clustering in sample size calculations, so we aim to provide methods that can be used to improve the design of such trials. An online app is coming soon!
This is work from the University of Adelaide, @sahmri.bsky.social and AusTriM. 4/4
This is work from the University of Adelaide, @sahmri.bsky.social and AusTriM. 4/4
July 28, 2025 at 8:22 AM
We know that many partially clustered trials do not account for clustering in sample size calculations, so we aim to provide methods that can be used to improve the design of such trials. An online app is coming soon!
This is work from the University of Adelaide, @sahmri.bsky.social and AusTriM. 4/4
This is work from the University of Adelaide, @sahmri.bsky.social and AusTriM. 4/4
We show that the DEFFs depend on the intracluster correlation coefficient (ICC), the range of cluster sizes, the distribution of cluster sizes, and the working correlation structure of the GEE analysis model. 3/4
July 28, 2025 at 8:22 AM
We show that the DEFFs depend on the intracluster correlation coefficient (ICC), the range of cluster sizes, the distribution of cluster sizes, and the working correlation structure of the GEE analysis model. 3/4
In this paper we derive design effects for trials with partial clustering due to pre-existing clusters (eg 👶 multiple births, 👀 eyes, 🦵 joints), where the clustered observations may be randomised at the cluster or individual level. 2/4
July 28, 2025 at 8:22 AM
In this paper we derive design effects for trials with partial clustering due to pre-existing clusters (eg 👶 multiple births, 👀 eyes, 🦵 joints), where the clustered observations may be randomised at the cluster or individual level. 2/4
Re terminology you might find our paper useful. We make the distinction between partially clustered/nested designs with pre-existing clusters and clusters formed after randomisation here: doi.org/10.1177/1740...
Sage Journals: Discover world-class research
Subscription and open access journals from Sage, the world's leading independent academic publisher.
doi.org
May 13, 2025 at 7:57 AM
Re terminology you might find our paper useful. We make the distinction between partially clustered/nested designs with pre-existing clusters and clusters formed after randomisation here: doi.org/10.1177/1740...
The partial clustering is often ignored when describing trials, in the sample size calculations, and in the analysis. More education and accessible tools are needed to improve the robustness of these trials. And that's what I'm currently working on - stay tuned! 2/2
December 11, 2024 at 6:12 AM
The partial clustering is often ignored when describing trials, in the sample size calculations, and in the analysis. More education and accessible tools are needed to improve the robustness of these trials. And that's what I'm currently working on - stay tuned! 2/2
A second paper compares the performance of mixed effects models and GEEs for partially clustered trials with clusters of size 2 (think twins 👶, eyes 👀, knees 🦵), including whether existing sample size formulas based on GEEs also provide appropriate power if analysis will be via mixed models: (2/3)
November 15, 2024 at 12:44 AM
A second paper compares the performance of mixed effects models and GEEs for partially clustered trials with clusters of size 2 (think twins 👶, eyes 👀, knees 🦵), including whether existing sample size formulas based on GEEs also provide appropriate power if analysis will be via mixed models: (2/3)