Linear Prediction

Linear prediction is the extrapolation of data points beyond what was collected in the time domain. This is done to improve resolution in the indirect dimensions and to avoid truncation artefacts. Linear prediction is therefore only done where the signal has not decayed to zero as otherwise it will result in additional noise with no improvement in resolution.

in general forwards linear prediction is done with an expanding time evolution whilst forward-backward (mirror-image) linear prediction is done with constant-time evolution in the indirect dimension.

NMRPipe manual pages for are available for linear prediction.