Skip to main content

Table 5 Comparisons between models constructed using the NIR spectra from ASD FieldSpec®3 and Polychromix Phazir™ with various wavelengths

From: A high-throughput quantification of resin and rubber contents in Parthenium argentatum using near-infrared (NIR) spectroscopy

 

DRY + IRR

DRY

IRR

\({\text{R}}_{\text{P}}^{2}\)

RMSEP

RMSEC

RMSEP/RMSEC

F

RPDP

\({\text{R}}_{\text{P}}^{2}\)

RMSEP

RMSEC

RMSEP/RMSEC

F

RPDP

\({\text{R}}_{\text{P}}^{2}\)

RMSEP

RMSEC

RMSEP/RMSEC

F

RPDP

Resin

 ASD (750–2500 nm)

0.764

1.486

1.437

1.034

6

2.056

0.829

1.200

1.070

1.121

7

2.413

0.765

1.343

1.297

1.036

7

2.065

 ASD (1100–2400 nm)

0.772

1.460

1.435

1.017

7

2.092

0.846

1.139

1.104

1.032

7

2.542

0.768

1.334

1.273

1.048

7

2.078

 ASD (1600–2400 nm)

0.758

1.503

1.478

1.017

7

2.032

0.837

1.170

1.140

1.026

7

2.474

0.742

1.407

1.351

1.042

7

1.971

 Polychromix (1600–2400 nm)

0.656

1.791

1.824

0.982

6

1.705

0.699

1.568

1.628

0.963

6

1.847

0.617

1.715

1.682

1.020

7

1.617

Rubber

 ASD (750–2500 nm)

0.756

0.753

0.731

1.030

7

2.024

0.780

0.724

0.661

1.095

7

2.128

0.755

0.647

0.618

1.047

6

2.019

 ASD (1100–2400 nm)

0.759

0.750

0.741

1.012

7

2.033

0.793

0.702

0.685

1.024

7

2.196

0.746

0.658

0.633

1.039

5

1.986

 ASD (1600-2400 nm)

0.720

0.808

0.805

1.004

5

1.887

0.752

0.769

0.750

1.025

6

2.005

0.757

0.644

0.612

1.053

7

2.029

 Polychromix (1600–2400 nm)

0.633

0.885

0.890

0.994

6

1.722

0.644

0.855

0.907

0.942

7

1.802

0.588

0.839

0.785

1.069

6

1.558

  1. All the models were constructed using the optimal combination of standard normal variate (SNV) + de-trending (DT) and Savitzky–Golay 2nd derivative under external validation (EV) context
  2. \({\text{R}}_{\text{P}}^{2}\), coefficient of determination; RMSEP, root mean squared error of prediction; RMSEC, root mean squared error of calibration; F, number of factors used in calibration; RPD, ratio of performance over prediction