Talk
When companies update a manufacturing process or replace certain materials, decision-makers face a critical question: Will the new version still meet quality standards and behave consistently? Traditional statistical tools like means and confidence intervals often fall short in providing a complete picture of variability, focusing only on average outcomes. This talk introduces statistical tolerance intervals as a powerful but underutilized tool for achieving faster, more reliable decision-making in change management.
We'll start with a clear explanation of the concept of tolerance intervals, comparing them with the more familiar confidence intervals. Using concrete analogies and visual graphics, audiences will understand how a tolerance interval essentially gives a safety margin range that is expected to capture a large majority (e.g., 90% or 99%) of future data points with high confidence. This is critical in industry settings when ensuring a new process or product remains within acceptable bounds for most units -- not just on average.
Last, we will dive into real-life inspired case study of a product change scenario using a non-linear model to represent a performance profile (such as the puff-per-puff output curve). We will demonstrate how to construct tolerance intervals for this model using R/Phyton, and how these intervals help confirm that a new design's output remains comparable to the reference's output range. Without delving into heavy mathematics, we will present a few essential formulas and pragmatic code snippets to illustrate the concept. Attendees will leave with practical insight into applying tolerance intervals, understanding when to use them over confidence intervals, and how this approach accelerates data-driven decisions in real-world projects.
About the Speaker
Dea Putri is a statistician with over a decade of experience bridging rigorous statistical analysis with practical decision-making. She has applied advanced modeling techniques across diverse regulated industries -- from pharmaceutical research to consumer product science -- tackling complex, real-world challenges in each domain. In her current role as a senior statistician at a global science & innovation company in Switzerland, Dea helps to shape statistical strategy for produce R&D and quality initiatives, working closely with scientists, engineers, and stakeholders to drive evidence-based improvements. She previously honed her skills in pharmaceutical R&D, collaborating with cross-functional teams on data-driven insights in early clinical trials.
Dea holds an M.Sc. in Statistics (Biostatistics) from Hasselt University in Belgium. A passionate advocate for statistical literacy, she strives to demystify advanced analytics for broad audiences and is motivated by her mission to connect rigorous quantitative thinking with real-world impact.
Dea's approachable style and cross-disciplinary perspective make her particularly skilled at translating complex statistical concepts into actionable insights for non-specialist teams, aligning perfectly with the PyData spirit of practical, inclusive data science.