During a parameter optimization, the effect on a calculated datum caused by changes in the parameters can be predicted. If the calculated change is very different to that predicted, the parameter optimization process might fail to converge. Most runs do not have any problems, and in a run that does have problems, normally only one or two data are involved. Rather than stop such a run, the weights of the offending data are automatically reduced. To prevent this automatic reduction in weights, add keyword NOSCALE. NOSCALE on the PARAM keyword line affects all reference data.
An alternative, that prevents the weights of individual reference data files from being changed, is to have the word NOSCALE on the third line of individual reference data files.
NOCALE does not affect normal data, that is, for normal data the default scaling faction is 1.0. For data that cause problems, when NOSCALE is not present the scaling factor decreases so as to reduce the impact of that data. If an individual datum stops causing problems, the scaling factor will increase to 1.0 again, for that datum.