DescriptionThe question of determining the relevant sample size is an important first step in the design of any statistical experiment. Choosing a too small sample size may not give enough statistical power to answer the research questions of interest, and choosing a too large sample may waste resources in terms of time or money. For instance, when planning to carry out a clinical study to test for the effect of a new drug, one clearly does not want to involve more patients than necessary, but having too few patients may not lead to signficant test results even if the drug in fact does have the effect under investigation.While at some occasion the sample size may be dictated by practical constraints (such as costs, or availability of samples), in general it would be desirable that, given a research question and a pre-determined power or accuracy level, one can calculate the required sample size for the study at hand. This project will investigate statistical methods to carry out these sample size calculations. As a starting point, we will consider sample size calculations for simple problems such as 2-sample t- tests or the estimation of population proportions [1]. The methods needed to carry out such calculations are relatively simple, and are implemented in many online tools such as [4]. We will study these methods, and then envisage to turn to more complex problems. The studied methods will be applied on practical problems as motivated by the work carried out for the design of actual clinical trials by the Durham Biostatistics unit (DBU).
PrerequisitesStatistics I
Resources
|
email: jochen.einbeck "at" durham.ac.uk