Greg Kiar, PhD
banner
gkiar.bsky.social
Greg Kiar, PhD
@gkiar.bsky.social
Research Scientist & Director of the Center for Data Analytics, Innovation, and Rigor @ The Child Mind Institute.

Computational medicine, uncertainty quantification, applied ML, software standardization, and some sports analytics for good measure.
📚 Read the full perspective here: www.nature.com/articles/s41...

... For a deeper dive into how embracing variation can advance neuroimaging research.

#SciComm #NeuroscienceResearch

8/8
Why experimental variation in neuroimaging should be embraced - Nature Communications
Brain imaging analysis lacks accessible ground-truth approaches, leading to varied results across the field. Embracing analytical variability may allow researchers to enhance the generalizability of f...
www.nature.com
November 14, 2024 at 1:04 AM
🧠 Overall, we argue for a paradigm shift in neuroimaging research:

Moving beyond mere reproducibility towards replicability, generalizability, and robustness.

7/8
November 14, 2024 at 1:04 AM
✅ We’ve made a checklist that we share through the #NMIND website to help researchers incorporate and report variability analysis in their studies.

This simple tool aims to standardize and encourage these practices.
nmind.org/variability-...

6/8
Variability Capture Checklist
NMIND (this Neuroimaging Method Is Not Duplicated) is a collaborative dedicated to accelerating scientific discovery in neuroimaging research that was formed in 2020 as a grassroots initiative- resp...
nmind.org
November 14, 2024 at 1:04 AM
While there are challenges like increased costs and complexity, these can be mitigated through open science practices and shared resources 🤝.

The long-term benefits outweigh the initial hurdles. #CollaborativeScience

5/8
November 14, 2024 at 1:04 AM
💡 Embracing variation offers several benefits:

It improves transparency, enhances generalizability, reduces the risk of p-hacking, and allows for quantification of result stability.

This leads to more robust scientific findings 🦾. #ReproducibleScience

4/8
November 14, 2024 at 1:04 AM
🔍 Each of these can significantly impact results, so our perspective is this

As much as possible, we should avoid making these decisions, and instead explore the multiverse-of-methods.

3/8
November 14, 2024 at 1:04 AM
🛠️ When you’re constructing a study, there are many decisions you make:

What data to use, experts to include, analytic options to explore, tools to use, systems to run on, and perturbations/contrasts to introduce.

2/8
November 14, 2024 at 1:04 AM
🚨New paper alert!🚨
Does the sheer number of available analysis workflows (and their potential for conflicting results!!) keep you up at night?

Our new 📄🔽 in Nature Comms explores how embracing this variability may actually improve generalizability of results.

1/8 🧵
November 14, 2024 at 1:04 AM