Nested data structures (common in neuroscience) create dependence in the data that influences the effective sample size and statistical power of a study. Several methods are available for dealing with nested data, including the summary statistics (SS) approach and multilevel modelling (MM). Recent publications have heralded MM as the best method for analysing nested data, claiming benefits in power over SS approaches (e.g. the t-test). However, when data are not cross-nested and cluster size is equal, conventional analysis with the SS approach is actually mathematically equivalent to MM. This equivalence has been established in statistical literature and made use of in popular neuroimaging software such as FMRIB Software Library (FSL) and Statistical Parametric Mapping (SPM). However, the merits of the SS approach have not been well recognised in neuroscience beyond that context. We conducted statistical simulations demonstrating equivalence of MM and SS approaches for analysing nested data and provide support for the utility of the conventional SS approach in nested experiments.
OSF does not support the use of Internet Explorer. For optimal performance, please switch to another browser.