BUCHI NIRCal - Manual (Page 160)

NIRCal 5.5 Software Manual
160 NIRCal 5.5 Manual, Version A
Row Name Description
82 Validity* The V-Set regression coefficient should be near to 1:
1-VsetRegr. Validity should be close to 0
83 Comparability* C- and V-Set should have similar high regression coefficient
of prediction:
Abs(CsetRegr-VsetRegr). Comparability should be close to 0
84 Precision* V-Set standard error of prediction (BIAS corrected) should be
low:
SEP/Abs (Range)
85 Weighted Accuracy* V-Set residual sum of squares=RSS should be low (not BIAS
corrected):
Abs(Sum((Orig-Predicted)^2)/Abs (Range)
* used with different weighting values.
Statistics of the C-Set
88
Property statistics Selected property
90 C-Set BIAS Always zero by definition.
91 C-Set SEE (SEC) Standard error of estimation of the C-Set. Should be as small as
possible, but comparable to the standard deviation of the
laboratory method.
92 CV SECV Standard error of estimation of the C-Set. Should be as small as
possible, but comparable to the standard deviation of the labor
method.
94 C-Set Regression
Coefficient "r" shows, how well the predicted values of C-Set match the
original values on average. Should be as close to 1 as possible (r
> 0.9)
95 C-Set Regression
Intercept r= a + bx, where "a"= intercept of the C-Set. Should be as close
to 0 as possible
96 C-Set Regression Slope r= a + bx, where "b"= slope of the C-Set. Should be as close to 1
as possible