A system is provided for processing data values in a data set to create a data value representation having a signal side. The data set contains valid data values and noise data values where the number of the valid data values is statistically insignificant with respect to the number of noise data values. The system determines a statistically significant portion of the signal side and fits a curve to the statistically significant portion which extends beyond the statistically significant portion of the signal side. Once the curve is determined, the system can determine a data-validity confidence value for one of the data values responsive to the curve. The data-validity confidence value can be used determine that a data value in a set of spectral data is not a noise data value.