Property f-test forward, Property interference, Property missing values – BUCHI NIRCal User Manual
Page 190: Property nearest neighbor, 58 property f-test forward, 59 property interference, 60 property missing values, 61 property nearest neighbor

NIRCal 5.5 Software Manual
190
NIRCal 5.5 Manual, Version A
3.18.58
Property F-Test forward
Description
Significance of additional PCs.
Use
For secondary principal component selection.
Method
PCR / PLS
Matrices ID
38
Tip
Details
Alternative for proofing the robustness of a calibration with a backward
search operation is to start at a certain calibration checking for additional
PCs that will improve the SEP value significantly.
Related Topic
Property F-Test backward
3.18.59
Property Interference
Description
Interference = Property Score Extension / Property Nearest Neighbor
Use
Judgement of the influence of the property for the cluster separation.
Method
Cluster (CLU)
Matrices ID
61
Tip
Small interference indicates a good influence of the property for the
calibration.
Details
Property Interference is the relationship between the size of a cluster and
distance to the nearest property.
The interference values are listed in the calibration protocol.
Related Topic
Property Score Extension
,
Property Nearest Neighbor
3.18.60
Property missing Values
Description
Spectra, which have no defined property value (0 = value missing; 1 = value
available)
NOTE
This has no relevance in NIRCal file-mode, only in database-mode.
Use
Find spectra, which have no property value
Method
PCR / PLS / Cluster (CLU) / SIMCA / MLR
Matrices ID
115
Tip
Edit the property values in Manager Console/ Sample Management
Details
Spectra without property values can not be used for the C- and V-Set.
Related Topic
Original Property
Interpretation:
1 means: there is defined property value,
0 means: missing property value = there is no defined property value.
3.18.61
Property Nearest Neighbor
Description
The smallest Mahalanobis distance between the property mean to the next
property center.
Use
Judge the property separation.
Method
Cluster (CLU)
Matrices ID
60
Tip
High values indicate good separation.
Details
These values together with the next neighbors are listed in the calibration
protocol.
Related Topic
Scores