Jackknife resampling

Jackknife resampling can be used to reduce the effect of outlier points in the time series. This is particularly useful when analysing real world datasets, which come without replicates. Jackknifing can also provide a useful measure of uncertainty around the estimated α value for each dataset.

To see whether Jackknifing improves the performance of this algorithm, for each dataset a number of subsamples is taken each containing 85% of the points in the original dataset. Points within the same subsample are sampled without replacement, however, the same point can be present in different subsamples. Each subsample is then processed as a normal dataset. The final α estimate is taken to be the average of the estimates of all subsamples derived from the same original dataset.