Good statistical work deserves to be understood. Here are my favorite strategies for making complex ideas accessible:
Instead of describing your model first, start with: “We wanted to know if…”. This keeps the focus on the goal, not the math.
Instead of “We fit a hierarchical logistic regression with shrinkage priors”
, say “We used a model that accounts for uncertainty across hospitals while still allowing us to compare them fairly.”
Most people think “uncertainty” means you're not sure. Frame it as what we know and how confident we are — show ranges, not just point estimates.
If your model uses latent classes, say: “We grouped patients into categories based on their symptom patterns — even if we don’t directly observe their disease.”
Replace tables with visual summaries. A clean bar plot with confidence intervals speaks volumes.
Always end with what your analysis means for decision-making, policy, or practice. Anticipate what a clinician, administrator, or policymaker might ask.