Abstract A widely used method for the analysis of data collected serially over time is to reduce the data on each subject to a single summary that reflects some salient aspect of the response. The summaries from all subjects are then analysed using appropriate univariate methods. In some cases it may be unreasonable to suppose that summaries in a group share a common distribution. A weighted analysis may be preferable, with data from each subject used to determine the weights. This approach is explored in the context of an analysis of data on foetal distress during labour.