# NUTS Help

## Linear Prediction (LN)

This command performs either forward or backward linear prediction, depending on parameters set by the user. Backward prediction can be used to correct corrupted early data points, which cause rolling baselines. Forward prediction is used to predict data out to twice the actual acquisition time, and is used with severely truncated data, such as in the indirect dimension of 2D experiments as an alternative to zero-filling. Show me backward or forward linear prediction.

When the LN command is issued, a dialog box is displayed prompting for the following values:

- Forward or backward prediction

Number of data points on which to base the prediction

Number of points to be predicted for back prediction

Maximum number of frequencies to predict.

For backward prediction (which is the default), the user must set the number of points to back predict, the number of data points upon which to base the prediction and the number of frequencies to predict. The default value for number of points to back predict is 2 and should usually be a small number (4 or less). Larger values may cause failure of the prediction algorithm. The number of data points on which the prediction is based must be less than half the total number of data points, or the algorithm fails. The larger this value, the longer the prediction process will take. The number of frequencies to predict is usually unknown, but usually a small value can be used. The larger the value, the higher the chances that the algorithm will fail. The option has been added of allowing the algorithm to determine the number of frequencies, by setting this parameter to -1.

For forward prediction, the user must select the Forward Prediction button at the bottom of the dialog box. Forward linear prediction always doubles the number of data points. The value for number of points to back predict is ignored.

LN can be used in Links and Macros, in which case the dialog box does not open. The parameters must be chosen before starting automated processing, and the last set of values will be used. It is advisable to experiment with parameters before initiating automated processing.

The values of all linear prediction parameters can be explicitly set in a macro, using SET commands, as shown in the examples below:

- SET LNpts = 4

SET LNmdim = 64

SET LNnsig = 16

SET Lndirection = FORWARD

SET Lndirection = BACKWARD

Linear prediction takes considerable time (can take 20-30 min for a 2D data set, even on a fast computer).

See also: Polynomial baseline correction

Last updated: 11/3/97.