Attending the Using Simulation Studies to Evaluate Statistical Methods course

I recently attended a one day course in using simulation to evaluate statistical methods. This course was run by the Institute of Clinical Trials and Methodology at UCL. The instructors were Dr. Tim Morris, Professor Ian White and Dr. Micheal Crowther.

The course was very hands on, with a team exercise to write code to use simulation to test the properties of the t-test with skewed or uneven sample sizes. The exercise was like a mini-dissertation, with having to narrow the focus of the question, and to write a simulation SAP, and to present results.

One of the main teaching points is that simulation studies need careful planning and a written analysis plan, just like clinical trial analysis. In addition, careful thinking about the purpose of the simulation, such as proof of concept, or comparison of various statistical methods was discussed. Another concept is the outcome measure of your simulation, and a bit of discussion about finding the appropriate number of simulations needed.

I would highly recommend this course, and I will be presenting a few ideas from the course in the CSM Internal Seminar on June 22.