Chen Research Group

Radiation Oncology Corner of Statistics

Welcome to Radiation Oncology's Corner of Statistics, within the Department of Radiation Oncology at the University of Kansas Medical Center.

We will present on a variety of topics in statistics, data analysis and applications. Hopefully, you will bookmark this webpage onto your PC or laptop for a regular visit.

Thank you and enjoy your time!

2025

Statistical topic 1 for radiation oncology: A Brief Introduction to Bayesian Data Analysis Presentation (Word)

In this topic, we present some basic concepts in Bayesian data analysis and provide an example to show how this popular method can give very similar results to those from traditional statistical methods.

Graph depicting the posterior depiction of beta0

Statistical topic 2 for radiation oncology: Statistical index variables and applications presentation (Word)

In this topic, we are introducing some popular and influential statistical index variables which are a comprehensive score of the underlying phenomena. At first sight, they are minor, yet they are powerful in the applications to wide range of clinical practice and research.

line graph depicting the JAMA magazine publications from 1999 to 2021

Statistical topic 3 for radiation oncology – Data transformations in statistical analysis presentation (Word)

In this topic, we present some SAS and R codes introducing fundamental methods in data transformation, either from wide (horizontal) data to long (vertical) formatted data or vice versa; and why we need these procedures in statistical analysis.

data transformation graph - tumor size over time by patient

Statistical topic 4 for radiation oncology at KUMC – Cross validations in statistical applications. Presentation (Word)

In this topic, we are introducing the concepts of cross validation (CV) – how it is defined, types of cross validation, why we need it in statistical analysis (key ideas behind CV), a brief historical note and some examples in R and SAS.

plot graph of predicted vs actual sepal length (bootstrap sample)

Statistical topic 5 for radiation oncology at KUMC – Some common mistakes in statistical analysis. Presentation (Word)

In this topic, we are talking about some commonly seen mistakes in research or study. Some are obvious, some are quite hidden – 2 Kaplan-Meier curves could overlap or cross over but still significantly different; a model looks great but may be very badly overfitted; correlations are not causations; we could visually exaggerate some minor group differences in the data; and more. R and SAS codes are provided to replicate the examples.

graph of overlapping/crossing KM Curves w/statistically signifidant p-value

Stat Archive

Statistical topic 1 for radiation oncology - A glimpse of medicine – a poem: Presentation

The poems tell stories about how data and statistical analysis affect the daily work of the data analyst at the Department of Radiation Oncology, KUMC.

Statistical topic 2 for radiation oncology at KUMC - Non-inferiority and superiority tests in clinical trials: Presentation

In this topic, we are introducing some of the fundamental concepts that are a key to understanding clinical trials where we would apply the non-inferiority or superiority statistical test to evaluate a new treatment or new drug and so on.

Statistical topic 3 for radiation oncology at KUMC - Why is logistic model so popular? Presentation

In this topic, we’ll introduce the basic concepts for performing logistic models and how they are realized in many different areas such as clinical data, banking service, NFL, and more.

Statistical topic 4 for radiation oncology at KUMC - Uncertainty in statistical applications: Presentation

In this topic, we present some types of uncertainties that appear in statistical analysis such as SD, bias, asymptotic theory, adaptive statistics in clinical trials, fuzz statistics, adjustments in statistics and quasi-statistics.

Statistical topic 5 for radiation oncology at KUMC – Geometric curves and their equations in statistical applications: Presentation

In this topic, we present some interesting statistical curves we see every day and their corresponding algebraic equations (a magic duality). Examples include normal distribution curve, S shaped curves, KM curves, logistic curves, and more.

Statistical topic 6 for radiation oncology at KUMC – Entropy, information theory and statistical applications: Presentation

In this topic, we tried to link the concept of entropy in physics (thermodynamics and statistical mechanics...) to information theory (computer science, communications...) and statistical applications. It remains mysterious how all these quite different branches of science are connected via a low-profile creative probabilistic idea.

Statistical topic 7 for radiation oncology at KUMC – Mistakes or misunderstanding in statistical applications? Presentation

Statistical applications are a part of the integrated solution of clinical practice, especially in writing a paper, a grant application, or a presentation at a professional meeting. Can we avoid any mistakes when we do statistics? Not likely.

Statistical topic 8 for radiation oncology at KUMC - How to select good variables for a multivariable model? Presentation

This is a common situation that occurs again and again when we were trying to model a clinical data or observational cohort. We have a larger number of variables in hand but how should we select some “good” variables (also nicknamed as “predictors”) so that we can use them for a multivariate predictive model such as logistic model or Cox proportional hazard model?

Statistical topic 9 for radiation oncology at KUMC - Nomogram and its applications. Presentations

A nomogram is an alternative to a mathematical formula in presenting a complex result of a data, for example a multivariable logistic regression or Cox model in survival analysis. We should try to include it in our data analysis for logistic model or Cox proportional model in survival data analysis. It would help us to visualize the data and the biological or clinical mechanism that may be hidden in the database.

Statistical topic 10 for radiation oncology at KUMC - Randomness, hovering like a Himalaya eagle in the world of data. At first sight, it is seemingly related to chaos and order of beautiful nature, but we are far away from grasping its grandness, versatile applications, and universality. Presentations

Statistical topic 11 for radiation oncology at KUMC - Additive models in statistical analysis. Why do we use additive models? It’s not new but other than a general linear model, additive models are more flexible to deal with non-linear data, better in prediction than a linear model, more appropriate to handle missing data issues, and easier to interpret on the model so why not! Presentations