The effect of serial-correlation structure of biomedical data on the statistical-sampling variability of power estimates is investigated, extending a previous study of mean value estimates. The degrees of freedom content for power DFp in a k-point sample is shown to be dependent on the serial-correlation structure in a generally different way from the DFk, the corresponding measure for the mean. The sampling distribution of power is also investigated. The influence of serial correlation structure on the behaviour of traditional small-sample statistical tests is also of interest; in the case of the t-test the effects can be severe but can be compensated, leading to a new variable, ct, which for correlated data, behaves rather like t for uncorrelated data. The approach is applied to the case of long-term records of systolic blood-pressure data.
Statistical variability of biomedical data: part 2. The influence of serial correlations on power estimates, and on comparative testing of samples.
RUGGIERO, CARMELINA;
1981-01-01
Abstract
The effect of serial-correlation structure of biomedical data on the statistical-sampling variability of power estimates is investigated, extending a previous study of mean value estimates. The degrees of freedom content for power DFp in a k-point sample is shown to be dependent on the serial-correlation structure in a generally different way from the DFk, the corresponding measure for the mean. The sampling distribution of power is also investigated. The influence of serial correlation structure on the behaviour of traditional small-sample statistical tests is also of interest; in the case of the t-test the effects can be severe but can be compensated, leading to a new variable, ct, which for correlated data, behaves rather like t for uncorrelated data. The approach is applied to the case of long-term records of systolic blood-pressure data.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.