Simulation Studies for Evaluating Statistical Models

Simulation studies offer a way of judging whether a statistical method works in the way that we think it should under specific circumstances. Even commonly used methods that we think are ‘tried and tested’, actually don’t work in the circumstances that we want to use them. It is therefore essential that we can understand the results of simulation studies, and also run simulations ourselves to ensure that we know how a statistical model should behave in the specific circumstances we want to use them. Simulations also help us to understand what the model we are using is really doing, which is helpful for ensuring the conclusions we are making are really justified.

This course will introduce researchers to the concepts that underlie simulation studies, their uses, and their limitations. It will also provide experience of running simulations in either R or Stata, so users will be able to run their own studies.

Participants should be competent at using either R or Stata, as the exercises will be conducted in these packages. They should have their chosen package installed and ready to use on their computer. Participants should also have a solid understanding of, at least, multiple regression models.

Outcomes

Participants in this course will:

• Understand what simulation studies are and what they can show;
• Learn how simulation studies work, and the measures used to judge the results of a simulation study;
• Learn how to use the R software to run simple simulation studies;
• Learn how to produce a range of plots (in R) to visualise the results of simulation studies.

Contributors

Dr Andrew Bell’s research spans a diverse range of subject areas: his work includes a focus on health inequalities, for example looking at mental health trajectories from a life-course perspective; but he has also contributed to other disciplines, including geography, political science, and economics. His work is united by a methodological interest in the development and application of multilevel models, with a focus on age-period-cohort analysis and fixed and random effects models. He is currently working on an ESRC-funded project developing the use of multilevel models for uncovering intersectional inequalities, particularly in health outcomes.

Dr Calum Webb’s research explores socioeconomic inequalities in the child welfare system and their relationship to fiscal and social policy using quantitative research methods. His research on child welfare inequalities and the funding of local services for children and young people has been published in the Journal of Epidemiology and Community Health, Children and Youth Services Review, the Journal of Mixed Methods Research, Child & Family Social Work, and elsewhere. Outside of academic circles, his work has been cited by the National Children’s Bureau, Ofsted, Children England, the British Association of Social Workers, the Independent Review of Children’s Social Care, the Department for Education, and other organisations.