A modified paired t-test for the analysis of plant trials with data autocorrelated in time

Sequential plant performance data, such as daily metal recovery, are shown to follow a first order autoregressive time series model. This has been used to modify the standard formula for the paired t-test in evaluating on-off trials of alternative operating conditions, such as a new flotation reagent or circuit configuration. The modified test is more powerful than the standard test, requiring less trials to reach a decision at a given level of confidence. The power increases with the value of the AR(1) autocorrelation coefficient. per gives formulae for the modified test, and for calculating the number of data pairs required to detect a given difference at a particular confidence level. An example of the use of the new test is given.