BUCHI NIRCal - Manual (Page 156)

NIRCal 5.5 Software Manual
156 NIRCal 5.5 Manual, Version A
Explanation of V-Set t-test:
Original property and the Predicted property are compared via a paired t-test to show if the results are
statistically equivalent.
where:

e is the mean residual (original – predicted property);

s d is the standard deviation of the residuals.
Example:
The two-
sided critical value of students’"t" e.g. at 5% significance level for n-1 degrees of freedom is
protocoled as V-Set t (5%,n-1) and the calculated t as V-Set t-value. If V-Set t (5%, n-1) is greater than
V-Set t-value, there is no evidence for a difference between the Original and Predicted results.
NIRCal estimates and protocols the significance level, where is no evidence for a difference between
the Original and Predicted results.
V-Set t-Test (n-1,2-tail) Confidence [%]: the higher the value the best the confidence of the V-Set
results.
See reference: [Anthony C. Moffat, Andrew D. Trafford, Roger D. Jee and Paul Graham. “Meeting the
International Conference on Harmonisation’s Guidelines on Validation of Analytical Procedures:
Quantification as exemplified by a near-
infrared reflectance assay of paracetamol in intact tablets.”,
The Royal Society of Chemistry, Analyst, 2000, 125,1341-1351].
NOTE

the V-Set spectra selection has a huge influence on the t-value;

the t-test reacts also negative on calibrations with very small V-Set BIAS. The t-test for the C-
Set has no sence, because the C-Set BIAS is per default 0. t = Abs( BIAS ) / ( SEP / Sqrt(n) )
Statistics of the C-Set and V-Set:
Row
Name Description
136-
151 Validation for C-Set
Spectra Name, number of spectrum, original, predicted property and the
difference (original minus predicted); an X for "extrapolation", in
case the value is outside the calibration range; residual value and
an X for "residual too big", in case the value is outside the allowed
range, spectra GUIDs
154-
163 Validation for V-Set
Spectra