science reveals its variability

A large scale international study in which he has participated CSIC has revealed that different analyzes on the same data yield different scientific results, highlighting the extent to which methodological decisions influence scientific results.

This “does not necessarily imply errors, but rather the existence of multiple possible methodological paths within accepted scientific standards.”

“This study does not question the validity of science, but rather provides evidence to better understand how conclusions are generated and how they can be strengthened. In the words of the authors, moving towards more reliable science involves not only replicating results, but also understanding the diversity of analytical paths that can lead to them,” he stressed.

“Analytical variability” in action

The research, published in ‘Nature’, has brought together 457 independent analysts from institutions around the world, who carried out 504 reanalyses of data from 100 previously published scientific studies. Everyone received the same set of data and the same research question, but had complete freedom to decide how to analyze it.

According to him CSICthe results demonstrate that factors such as the way of cleaning the data, defining variables, choosing statistical models or interpreting the results can significantly influence the final conclusions. This phenomenon is known as “analytical variability.”

The CSIC researcher at the Cajal Neuroscience Center (CNC), Guiomar Niso, carried out an advanced analysis of complex data with a rigorous and reproducible methodology, a strategic line that also characterizes her research in neuroimaging and open science.

“Each analysis involves methodological decisions that can influence the results. Understanding and quantifying this variability and guaranteeing transparency in analytical processes allows us to strengthen confidence in scientific research,” said Niso.

The research is part of the international program SCORE (Systematizing Confidence in Open Research and Evidence)designed to evaluate the robustness, reproducibility, and credibility of research in the social and behavioral sciences. According to the CSIC, Niso’s participation “reinforces the commitment” of the institution “to open science, methodological transparency and the development of standards that allow increasing the robustness of scientific results.”