Figure 2.

Resampling can be used to estimate the effect of measurement errors in SFX data on final refined coordinates. (a) Jackknifing-type resampling, see Nass Kovacs et al. (2019 ▸). Multiple resampled datasets are constructed by randomly drawing images from the entire original pool of diffraction images, resulting in resampled datasets that are smaller (70–90%) than the original dataset. Structures are determined from each of these, which are then compared to obtain an estimate of the variation in the atomic positions. (b) Bootstrap [see Grünbein, Foucar et al. (2021 ▸)] resampling is similar to jackknifing, but the resampling is performed by ‘random drawing with replacement’, which means that after random selection of an image from the pool a copy is placed in the resampled dataset and the original image is put back in the original pool. In this way, multiple resampled datasets are constructed that contain the same number of images as the original pool but in which images can be represented multiple times.