Interpretation of the Calibration Statistics

In our WebShop the Calibration statistics are shown as (example):

Calibration Set (36)
R2      = 0.98724
RPD     = 8.8525
RMSEC   = 0.3308
SEC     = 0.3355
Bias    = 0.0000

Test Set (29)
R2      = 0.98792
RPD     = 9.0970
RMSEP   = 0.3249
SEP     = 0.3304
Bias    = 0.0140

What does that mean?

The data is splited into 3 independent sets.
The “Reference vs. Predicted” plot shows the 3 sets in different colors.

The Calibration Set (CSet) to build the calibration and
the Validation Set (VSet) used to determine the calibration parameters and
the Test Set (TSet) used to measure the calibration performance.

The statistical values are listed for the CSet and the TSet as follow:

Calibration Set (the number of spectra per set)

R       = Poor 0.0 - 1.0 Excellent
          correlation coefficient or coefficient of correlation, 
          how close the data are to the fitted regression line. 

R2      = Poor 0.0 - 1.0 Excellent
          R-squared value or coefficient of determination. 
          R2 = R * R :  determination (R2) is the square of the correlation (R).

RPD     = Degrees of merit for the Ratio of Performance Deviation (RPD) to the application of NIR spectroscopy.

          RPD value       Rating              NIR Application

          0.0 - 1.99      Very poor           Not recommended
          2.0 - 2.49      Poor                Rough screening
          2.5 - 2.99      Fair                Screening
          3.0 - 3.49      Good                Quality control
          3.5 - 4.09      Very good           Process control
          4.1 -  oo       Excellent           Any application

RMSEC   = Accuracy      = total error       : Root Mean Square Error of Calibration 
SEC     = Precision     = random error      : Standard Error of Calibration = Sdev(x-y): as small as possible (around the standard deviation of the reference method)
Bias    = Trueness      = systematic error  : by definition 0 for the calibration.

Test Set (the number of spectra per set)

R       = Correlation   , as above
R2      = Determination , as above
RPD     = Applicativity , as above

RMSEP   = Accuracy      = total error       : Root Mean Square Error of Prediction
SEP     = Precision     = random error      : Standard Error of Prediction = Sdev(x-y) : as small as possible (around the standard deviation of the reference method)
Bias    = Trueness      = systematic error  : around 0

Total simplified

look at 
    Test Set  RPD       : for rating and applicativity
    Test Set  RMSEP     : for comparison with reference method

Digitization in the field of NIR spectroscopy (smart sensors)

Digitalization is advancing, also in NIR spectroscopy, which enables trainable miniature smart sensors e.g. for analyses in the food&feed, chemical and pharmaceutical sectors.

The calibration is the core of a NIR spectroscopy sensor, it enables the numerous applications and should therefore not be the weakest link in the measurement chain.

The development of calibrations that turn NIR spectrometers into smart sensors is done manually by experts (NIR specialist, chemometrician, data scientist) with so-called chemometrics software.

This is very time-consuming (time to market) and the result is person-dependent and thus suboptimal, because each expert has his own preferred way of proceeding. In addition, the calibrations have to be maintained, as new data has been collected in the meantime, which can be used to extend and improve the calibrations.

This is where our automated service comes in, combining the knowledge and good practices of NIR spectroscopy and chemometrics collected in one software and using machine learning to generate optimal calibrations.

Based on this, we have developed a complete technology platform (Time to Market) that covers the entire process from sending NIR + Lab data, to NIR Calibration as a Service, from online purchase of calibrations, to NIR Predictor software that directly evaluates newly measured NIR data locally and generates result reports.

Besides the free desktop version with user interface, the NIR Predictor can also be integrated (OEM). This can be integrated in parallel as a complement to your current Predictor, allowing the user to choose how they want to calibrate. And give them the advantage in NIR feasibility studies and NIR spectrometer evaluations to quickly provide the customer with a solid and accurate calibration that will make their NIR system deliver better results.

Advantages for your NIR users (internal or external)
  • no initial costs (no chemometrics software license required),
  • calculable operating costs (fixed amount instead of time and hourly rate) (calibration development, calibration maintenance)
  • easy to use (no chemometrics and software training),
  • quicker to use (no calibration development work) and
  • better calibrations (precision, accuracy, robustness, …)

Our chargeable service is based on the calibration development and the annual calibration use. Calibration development and calibration use can also be carried out separately (manufacturer / user).

For you as a spectrometer manufacturer, this means that you can deliver your system pre-calibrated for certain applications without incurring software license costs. And without your application specialists having to provide additional calibration services.

The unique advantages of our calibration service together with the free NIR Predictor are:
  • no software license costs (chemometrics software, predictor software, OEM integration)
  • no chemometrics know-how necessary
  • no time needed to develop optimal NIR calibrations.

If interested in using/evaluating the service :

About : Time and knowledge intensive creation and optimization of chemometric evaluation methods for spectrometers as a service to enable more accurate analysis and measurement results.

see also

Paradigm Change in NIR

Five Mistakes to avoid on Digitalization in NIR

NIR – Total cost of ownership (TCO)

OEM / White Label Software

White Paper

NIR Instruments

The NIR-Analysis (NIRA) also known as Near Infrared Reflectance (NIRS-Analysis) or NIR Transmission (NIT-Analysis) uses so called NIR-Spectrometer (see also NIR-Spectroscopy, NIR Spectrometry).

The supported NIR-instruments, NIR-analysers, NIR-sensors and NIR-spectrometer (near-infrared spectroscopy) are full range NIR (780-2500 [nm] or 12’820-4’000 [1/cm]) from any manufacturer and technology and also Short Wave InfraRed (SWIR) (900-1700 [nm]), that is typically used in Hyperspectral Imaging (HSI) or VIS-NIR (400-2500 [nm]) or UV-VIS-NIR (200-2500 [nm]). Definitions by IUPAC

The supported NIR-technology can be Fourier transform (FT-NIR), dispersive NIR (DLP, MEMS), NIR-diode-array, Acousto-optic tunable filter (AOTF, AOTFNIR), etc. (on-line, in-line or at-line)

Time-saving Calibration Support for all NIRS

New : NIR-Predictor Software for all NIR spectrometers! Analyze your samples.

Start Calibrate

Example overview list of NIR-spectrometer manufacturers / vendors / brands / supplier (we DO NOT SELL instruments) :

New : NIR-Predictor Software for all NIR spectrometers! Analyze your samples.

Start Calibrate

Example overview list of Miniature Near-Infrared (NIR) Spectrometer Engine (spectral sensor) manufacturers / vendors / brands (we DO NOT SELL instruments) :

Disclaimer: We have no affiliation with any of these sites or their companies.
All trademarks belong to their respective owners and are used for information only.
We DO NOT SELL instruments.