摘要：Psychologists estimate the precision of their statistics both to conduct hypothesis tests and to construct confidence intervals. The methods traditionally used for this are available only for a small set of statistics (e.g., the mean and transformations of it) and often make unrealistic assumptions about the variables' distributions. These assumptions are often particularly unrealistic in data derived from clinical samples, or when looking at groups responding at the extreme end of clinical constructs. Bootstrap estimation is a computer intensive procedure that offers a flexible and automatic alternative. The computer takes thousands of bootstrap samples from the observed data and from these bootstrap samples estimates the precision of the statistic. High-speed personal computers make the bootstrap a viable and appealing technique throughout the sciences. This article offers a tutorial on the theory and practice of applying bootstrap estimation to data from clinical samples and measures relevant to experimental psychopathology.
