Chemometrics and Intelligent Laboratory Systems 110 (2012) 129–134

Contents lists available at SciVerse ScienceDirect

Chemometrics and Intelligent Laboratory Systems

j ourna l homepage: www.e lsev ie r .com/ locate /chemolab

The chemometrics approach applied to FTIR spectral data for the analysis of rice branoil in extra virgin olive oil

Abdul Rohman a,b,⁎, Yaakob B. Che Man c

a Laboratory of Analytical Chemistry, Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Gadjah Mada University, Yogyakarta, Indonesiab Research Center of Halal Products, Gadjah Mada University, Yogyakarta, Indonesiac Laboratory Analysis and Authentication, Halal Products Research Institute, Universiti Putra Malaysia 43400 Selangor, Malaysia

⁎ Corresponding author: Tel.: +62 274 543120; fax:E-mail address: abdulkimfar@gmail.com (A. Rohman

0169-7439/$ – see front matter © 2011 Elsevier B.V. Alldoi:10.1016/j.chemolab.2011.10.010

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 April 2011Received in revised form 12 October 2011Accepted 18 October 2011Available online 25 October 2011

Keywords:ChemometricsFTIR spectroscopyRice bran oilExtra virgin olive oil

Among eleven studied vegetable oils, rice bran oil (RBO) has the close similarity to extra virgin olive oil(EVOO) in terms of FTIR spectra, as shown in the score plot of first and second principal components. Thepeak intensities at 18 frequency regions were used as matrix variables in principal component analysis(PCA). Consequently, the presence of RBO in EVOO is difficult to detect. This study aimed to use the chemo-metrics approach, namely discriminant analysis (DA) and multivariate calibrations of partial least square andprinciple component regression to analyze RBO in EVOO. DA was used for the classification of EVOO andEVOO mixed with RBO. Multivariate calibrations were exploited for the quantification of RBO in EVOO. Thecombined frequency regions of 1200–900 and 3020–3000 cm−1 were used for such analysis. The resultsshowed that no misclassification was reported for the classification of EVOO and EVOO mixed with RBO. Par-tial least square regression either using normal or first derivative FTIR spectra can be successfully used for thequantification of RBO in EVOO. In addition, analysis of fatty acid composition can complement the resultsobtained from FTIR spectral data.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Today, the chemometrics techniques have played a very importantrole in the study of edible fats and oils, especially for the authentica-tion study [1]. One of the chemometrics techniques widely used ismultivariate calibrations in order to elaborate the relationship be-tween the concentration of analyte(s) and the response of instrumen-tal assay like FTIR spectra [2]. The chemical analysis by infraredspectrophotometry rests on the fast acquisition of a great number(several hundred and even several thousands) of spectral data [3].

Fourier transform infrared (FTIR) spectroscopy has emerged an at-tractive alternative technique for some reasons. The development ofattenuated total reflectance (ATR) as sampling handling techniquehas revitalized the use of FTIR spectroscopy. Using ATR, there is no ex-cessive sample preparation; consequently, the use of hazardous sol-vents and reagents can be avoided [4]. This fact is very attractive forscientists who are take care about the human and environmentalhealth issues. For this reason, FTIR spectroscopy and other vibrationalspectroscopic techniques can be taken into consideration as “greenanalytical technique” for the analysis of edible fats and oils [5].

In recent years, olive oil (OO) has received great attention owingto its biological activities and sensory qualities. It has social and

+62 274 543120.).

rights reserved.

economical importance for the Mediterranean regions [6]. OO is oneof the strictly regulated oil products; consequently, it can be targetfor adulteration. Among OO classes, extra virgin olive oil (EVOO) isthe highest quality of OO. Due to the therapeutic value and highprice of EVOO, some market players intentionally or unintentionallytry to blend EVOO with much cheaper plant oils like palm, soya, andsunflower oils [7].

The adulteration of food products is of primary importance forconsumers, food processors, regulatory bodies, and industries [8].The adulteration practice frequently involves the replacement or dilu-tion of high-cost ingredients with cheaper substitutes. Although theadulteration is done for economic reasons, the action can cause severehealth and safety problems such as the Spanish toxic syndrome thatkilled some people. In addition, the adulteration of EVOO can be a po-tential risk for patients having the allergic history to EVOO's adulter-ants [9, 10].

Several publications have reported the application of chemo-metrics techniques applied to FTIR spectral data for quantitative anal-ysis of certain plant oils. The presence of hazelnut oil [11], sunflowerand corn oils [12], sunflower, corn, soybean and hazelnut oils [13],sunflower, soybean, sesame, and corn oils [14], and palm oil [15]has been analyzed using FTIR spectroscopy combined with chemo-metrics techniques. However, there is no reported work in relationto the use of FTIR spectroscopy for the analysis of RBO which hasthe similar FTIR spectra to EVOO. The objective of this research wasto use the chemometrics techniques of discriminant analysis and

Fig. 1. The score plot of principal component analysis (PCA) for the differentiation of EVOO and other plant oils.

130 A. Rohman, Y.B.C. Man / Chemometrics and Intelligent Laboratory Systems 110 (2012) 129–134

multivariate calibrations (partial least square and principle compo-nent regressions) for the analysis of RBO in EVOO. Furthermore, thechange of fatty acid profiles in EVOO due to the addition of RBO wasalso reported in order to complement the FTIR spectroscopy results.

2. Experimental

2.1. Materials

Extra virgin olive oil (EVOO), rice bran (RBO), canola, corn, grapeseed, palm, pumpkin seed, soybean, sesame, sunflower, and walnutoils (WO) were purchased from the local market in Serdang, Selangor,

Fig. 2. The loading plot for the projection of PC1 and

Malaysia. The oil samples were packaged in polyethylene terephthal-ate (PET) bottles and the dates of manufacturing were not known.Virgin coconut oil (VCO) was obtained from Jogjakarta, Indonesiawith the brand name of POVCO®. The used VCO was made usingcold extraction under the supervision of Prof. Bambang Setiadji fromthe Department of Chemistry, Gadjah Mada University, Yogyakarta,Indonesia.

2.2. Classification

Classification of pure EVOO and EVOO adulteratedwith RBOwas car-ried out using discriminant analysis (DA) by computing theMahalanobis

PC2 using peak absorbancies as matrix variables.

Fig. 3. FTIR spectra of extra virgin olive oil and other vegetable oils as oil adulterants at frequency regions of 4000–650 cm−1.

131A. Rohman, Y.B.C. Man / Chemometrics and Intelligent Laboratory Systems 110 (2012) 129–134

distance from each class center of analyte(s) being classified. DA can beused to determine the class of RBO having the FTIR spectral similarityto EVOO [16]. To carry out DA, EVOO and RBOweremixed to obtain a se-ries of standard or trained sets of 20 pure and 20 adulterated samplescontaining 1–50% of RBO in chloroform. The use of chloroform in thisstudy is to facilitate the ease homogenization between RBO and EVOO.The samples containing RBOwere assigned as adulterated, while a seriesof pure EVOO in chloroform (5–100%) was marked with EVOO and clas-sified using FTIR spectra at selected frequency regions. All samples weremeasured using FTIR spectrometer.

2.3. Quantification

Quantification of RBO in EVOOwas performed using twomultivar-iate calibrations of partial least square regression (PLSR) and princi-ple component regression (PCR). Both calibrations transform theoriginal variables (FTIR spectra absorbancies) into the new ones,which are linear combination of original variables, known as factors[17]. PLSR and PCR are sometimes called with factor analysis. Bothtechniques relied on two steps, namely calibration and validation/prediction steps. In the calibration step, a mathematical model wasbuilt to correlate between the matrix of FTIR spectra (predictor) andthe concentration of analyte(s) of interest (response) from the refer-ence values. In the prediction step, the developed calibration modelwas used to calculate the concentration of unknown samples [18].For calibration, a set of 18 samples containing EVOO and RBOwas mixed together in the concentration range of 1.0–50.0% (v/v) ofRBO in EVOO. These samples were shaken vigorously to ensure the

Fig. 4. Details of the FTIR spectra of EV

total homogenization. For prediction or validation data set, another18 independent samples were built. The pure EVOO, pure RBO, andtheir binary mixtures were further analyzed using FTIR spectrometer.

2.4. FTIR spectra measurement and analysis

The spectra measurement of all samples was performed using FTIRspectrometer (Nicolet 6700 from Thermo Nicolet Corp., Madison, WI)equipped with a deuterated triglycine sulphate (DTGS) as a detectorand a KBr/Germanium as beam splitter. The instrumentwas interfacedto computer operating under Windows-based, and connected to soft-ware of the OMNIC operating system (Version 7.0 Thermo Nicolet).The rest of the procedure and condition were as previously reported[19].

2.5. Chemometrics analysis

Principal component analysis of FTIR spectra using absorbencies at18 wavenumber regions was carried out using the Unscrambler soft-ware (Camo, Oslo, Norway, USA). DA and multivariate calibrations(PLSR and PCR) were performed using the software TQ AnalystTM ver-sion 6 (Thermo electron Corporation, Madison, WI). The spectral re-gions where the variations were observed were chosen for developingPLSR and PCR as well as for DA. The optimum number of PLSR andPCR factors was determined using cross validation by plotting the num-ber of factors against the root mean square error of cross validation(RMSECV) and determining the minimum factors. The predictability

OO and RBO at 1200–900 cm−1.

Fig. 5. The Cooman plot for the classification of extra virgin olive oil (EVOO) and adulterated EVOO with rice bran oil or RBO (F).

132 A. Rohman, Y.B.C. Man / Chemometrics and Intelligent Laboratory Systems 110 (2012) 129–134

of the models was tested by computing root mean square error of pre-diction (RMSEP) as used by [21].

2.6. Fatty acid analysis

In order to determine the FA changes during EVOO adulterationwith RBO, EVOO was mixed with RBO in the range of 5–60% (v/v).These mixtures were kept in controlled room temperature (20 °C) be-fore being used for analysis. Determination of FA compositions in allsamples was carried out using GC-FID as reported in [20]. StandardFAMEs of 37 compounds (C4 to C24) (Sigma Chemicals, St. Louis,MO, USA) were used to identify the retention times. Quantitativeanalysis of FA was performed using internal normalization technique.

Percentage ð%Þ fatty acid x ¼ Peak area of fatty acid xTotal peak area of all fatty acids

� 100:

Meanwhile, FA changes during adulterationwere subjected to one-way ANOVA (analysis of variance) followed by Duncan multiple com-parison using SPSS version 17.0 software (SPSS Inc., Chicago, IL, USA).The significance value (p) less than 0.05 was statistically different.

3. Results and discussion

In order to know the plant oils having the close similarity withextra virgin olive oil (EVOO) in terms of FTIR spectra, the chemo-metrics of principal component analysis (PCA) was used. Fig. 1 exhib-ited the PCA score plot obtained from the correlation matrix of peakabsorbancies at 18 frequency regions, namely 3007, 2953, 2922,2853, 1743, 1654, 1463, 1417, 1402, 1377, 1236, 1160, 1117, 1098,1030, 962, 850, and 721 cm−1 (Fig. 3). In PCA, the first principal com-ponent (PC1) and the second principal component (PC2) account thelargest and the next largest of variable variation. PC 1 explained 72%variance, meanwhile PC 2 accounted 16%; therefore, an approximateof 88% of variance can be described by the first two PCs.

Fig. 1 described that among the studied plant oils, rice bran oil(RBO) has the closer distance to EVOO than others. EVOO and RBOwere separated on positive side, either in PC1 or in PC2. From theloading plot, the selected frequencies at fingerprint regions (1500–721) were more contributed than others, namely at 1117 and1236 cm−1 (Fig. 2). This fact supported that FTIR spectroscopy is afingerprint technique in which the samples can be differentiated in

Table 1The multivariate calibrations (PLSR and PCR) along with FTIR spectral treatments for quant

Regression Treatment Factor Equation

Calibration Predicti

PLS Normal 2 y=0.977x+0.663 y=0.991st der 5 y=0.998x+0.040 y=0.882nd der 4 y=0.982x+0.360 y=0.04

PCR Normal 10 y=0.999x+0.013 y=0.941st der 10 y=0.994x+0.111 y=0.802nd der 10 y=0.942x+1.159 y=−0

terms of the number of peak and the peak intensities at fingerprintregion [22].

3.1. FTIR spectra analysis

Fig. 3 exhibits FTIR spectra of EVOO, RBO, and other plant oils inmid infrared region (4000–650 cm−1). All spectra look very similarbecause all plant oils are mainly composed from triacylglycerols(90–95%) along with the di- and monoacylglycerols with the minorconcentration (about 5%) and other trace levels of some components.

Focusing on EVOO and RBO, there are minor differences betweenboth oils at a more detailed analysis in terms of small band shiftsand of small changes in the relative peak intensity (absorbancies), es-pecially at frequency region of about 3007 cm−1 and 1117 cm−1

(Fig. 4). The band at 3007 cm−1 was attributed from the stretchingvibration of cis-double bonds, meanwhile peak at 1117 cm−1 corre-sponds to the C–O group vibration [13]. These minor differencesserved as frequency region selection for the classification and quanti-fication of RBO in EVOO, as discussed in the following.

3.2. Classification

EVOO and EVOO mixed with RBO was classified using discrimi-nant analysis at the frequency regions of 1200–900 and 3020–3000 cm−1. The selection of the frequency regions is based on itsability to provide the classification power with no misclassificationresults between two classes (EVOO and EVOO adulterated withRBO). In addition, using detailed investigation, it is known thatthese frequency regions reveal the peak intensity differences be-tween EVOO and RBO.

Fig. 5 shows the Coomans plot for the classification of both classes.The x-axis shows the Mahalanobis distance to EVOO, while the y-axisshows the distance to EVOO adulterated with RBO. DA can classifypure EVOO and EVOO adulterated with RBO with accuracy level of100%. This means that no samples were mistakenly classified intothe wrong class.

3.3. Quantification of RBO in EVOO

Quantification of RBO in EVOO was carried out using multivariatecalibrations of PLSR and PCR. PLSR and PCR are themost used regressiontechniques in chemometrics [21]. In the optimization of frequency

ification of rice bran oil (RBO) in extra virgin olive oil (EVOO).

R2 RMSEC(% v/v)

RMSEP(% v/v)

on Calibration Prediction

0x−0.111 0.993 0.981 1.34 2.150x+1.560 0.998 0.931 0.620 2.455x+13.58 0.982 0.001 1.90 15.06x+0.490 0.999 0.962 0.380 1.803x+3.088 0.995 0.908 1.05 2.99.079x+17.47 0.942 0.003 3.40 16.3

Fig. 6. PLS model for the relationship between actual value and FTIR-predicted value ofrice bran oil using FTIR normal spectra at 1200–900 and 3020–3000 cm−1. A = calibra-tion; B = validation.

Table2

Thefattyacid

compo

sition

ofex

travirgin

oliveoil(

EVOO)ad

ulteratedwithrice

bran

oil(

RBO).

FAco

mpo

sition

(%w/w

)†Ra

tio(R

BO:EV

OO,v

/v)

(0%:10

0%)

(5%:95

%)(1

0%:90%

)(1

5%:85%

)(2

0%:80%

)(2

5%:75%

)(3

0%:70%

)(4

0%:60%

)(5

0%:50%

)(1

00%:0%

)

C14:0

0.01

±0.00

a0.09

±0.00

b0.11

±0.00

c0.14

±0.00

d0.15

±0.00

d0.17

±0.01

e0.19

±0.01

f0.28

±0.00

g0.29

±0.01

g0.36

±0.01

h

C16:0

10.83±

0.41

a12

.72±

0.20

b12

.97±

0.25

b13

.71±

0.02

c13

.81±

0.09

c13

.99±

0.10

cd13

.74±

0.07

c14

.34±

0.22

d16

.94±

0.12

e19

.08±

0.17

f

C16:1

0.76

±0.05

a0.61

±0.01

b0.59

±0.01

bc

0.56

±0.01

cd0.53

±0.02

d0.55

±0.01

cd0.52

±0.01

bd0.47

±0.01

e0.33

±0.01

f0.19

±0.01

g

C18:0

3.24

±0.14

a3.01

±0.04

b2.94

±0.02

b2.53

±0.10

c2.54

±0.05

c2.42

±0.03

cd2.42

±0.04

cd2.37

±0.04

cd2.29

±0.01

d1.93

±0.06

e

C18:1

73.27±

0.76

a66

.29±

1.04

b65

.01±

0.06

bc

65.07±

0.12

bc

62.42±

2.06

d62

.07±

1.05

d62

.70±

0.44

bc

59.20±

0.08

e49

.67±

0.09

f41

.49±

0.11

g

C18:2

7.06

±0.02

a13

.28±

0.09

b14

.13±

0.10

b16

.15±

0.16

c16

.58±

0.21

cd17

.11±

0.02

cd17

.65±

0.23

cd19

.76±

0.30

e26

.04±

1.02

f32

.38±

0.38

g

C20:0

0.60

±0.00

a0.79

±0.02

b0.81

±0.01

bc

0.89

±0.01

cde

0.91

±0.03

cde

0.88

±0.01

bcd

0.94

±0.01

bcd

e0.99

±0.02

e1.16

±0.05

f1.32

±0.11

g

C18:3

0.33

±0.03

a0.50

±0.00

b0.51

±0.01

db0.53

±0.05

db

0.54

±0.00

b0.54

±0.01

b0.55

±0.00

b0.62

±0.01

c0.67

2±

0.01

c0.76

±0.02

d

C20:1

0.29

±0.01

a0.31

±0.01

ab0.31

±0.00

abc

0.33

±0.02

bcd

0.33

±0.00

cd0.34

±0.00

de

0.34

±0.01

de

0.36

±0.00

e0.41

±0.00

f0.46

±0.01

g

C22:0

0.12

±0.01

a0.14

±0.01

ab0.14

±0.01

bc

0.16

±0.01

c0.16

±0.00

cd0.16

±0.00

cd0.16

±0.00

cd0.17

±0.01

d0.21

±0.00

e0.24

±0.01

f

FA=

fattyacid;†Ea

chva

luein

thetablerepresen

tsthemea

nsof

triplic

atean

alysis;SD

isgive

nafter±.M

eans

withinea

chrow

withdifferen

tlettersaresign

ificantly

differen

tat

Pb0.05

.C14

:0,m

yristicacid;C1

6:0,

palm

itic

acid;C1

6:1,

palm

itoleicacid;C1

8:0,

stea

ricacid;C1

8:1,

oleicacid;C1

8:2,

linoleicacid;C1

8:3,

linolen

icacid;C2

0:0,

decano

icacid;C2

0:1,

deceno

icacid;an

dC2

2:0,

dode

cano

icacid.

133A. Rohman, Y.B.C. Man / Chemometrics and Intelligent Laboratory Systems 110 (2012) 129–134

regions used, the regions containing the significant information wereselected and all the useless signals coming from the interferences or in-strumental drifts were ignored [18]. The frequency regions used forsuch quantificationwere based on its capability to provide the high cor-relation between actual and FTIR-predicted levels of RBO in EVOO.Based on this optimization, the frequency regions used for classification(the frequency regions of 1200–900 and 3020–3000 cm−1) were usedfor the quantification of RBO in EVOO.

Table 1 lists the performance of multivariate calibrations (PLSRand PCR) along with the FTIR spectral treatments (normal andSavitzy–Golay first and second derivatives) in terms of coefficient ofdetermination (R2), root mean square error of calibration (RMSEC),and root mean square error of prediction (RMSEP). In general, PLSRoffers the better results than PCR for quantitative analysis of RBO inEVOO. Furthermore, both normal and first derivative FTIR spectraoffer good model for such quantification. From Table 1, it is knownthat normal FTIR spectra give the better prediction model than firstderivative spectra, as indicated by RMSEP values (2.15% for normalspectra and 2.45% for first derivative). Inversely, the first derivativespectra give the better calibration model than normal spectra, as indi-cated by RMSEC values (0.620% for first derivative and 1.34% for nor-mal spectra). However, the differences of RMSEC and RMSEP valuesfor both normal and first derivative FTIR spectra are relatively low.

Using PLSR and normal spectra, the R2 values obtained are 0.993(in calibration) and 0.981 (in prediction). Two and five latent vari-ables (factors) were selected for building PLSR in normal and first de-rivative FTIR spectra, respectively. The high value of R2 and the lowvalue of errors in calibration and prediction indicated that FTIR spec-tral data combined with PLSR can be effective tools in terms of its ac-curacy and precision to measure the levels of RBO in EVOO.

Fig. 6 exhibits the scatter plot for the relationship between actualvalue (x-axis) and FTIR-predicted value (y-axis) of RBO in EVOO usingPLSR with normal FTIR spectra, showing the close relationshipbetween two variables assessed. This result indicated that FTIRspectroscopy combined with PLSR can be reliable technique for thequantification of RBO in EVOO.

Fig. 7. The relationship between the increasing level of rice bran oil (RBO) and the fatty acid changes.

134 A. Rohman, Y.B.C. Man / Chemometrics and Intelligent Laboratory Systems 110 (2012) 129–134

3.4. Fatty acid analysis

Analysis of fatty acid composition of edible oils seems to be an im-portant tool to detect the presence of specific oil (RBO) in another(EVOO). Table 2 shows the major fatty acids present in RBO, EVOO,and their mixtures. Palmitic (C16:0), stearic (C18:0), oleic (C18:1),and linoleic acids (C18:2) are the main fatty acids composing EVOOand RBO. Therefore, these fatty acids are used for the detection ofRBO in EVOO.

During the addition of RBO into EVOO, the levels of stearic andoleic acids were decreased linearly with increasing levels of RBOwith R2 values of 0.793 and 859, respectively. In addition, the levelsof stearic and linoleic acids were increased with the increasing con-centrations of RBO with R2 values of 0.823 and 0.884, respectively(Fig. 7).

4. Conclusion

The chemometrics approach of discriminant analysis (DA) andmultivariate calibrations of PLSR can facilitate the classification andquantification of RBO in EVOO at the combined frequency regions of1200–900 and 3020–3000 cm−1

. DA can discriminate EVOO andEVOO adulterated with RBO, with accuracy level of 100%. PLSR, eitherusing normal or first derivative FTIR spectra, offers reliable techniquefor the quantification of RBO in EVOO, as indicated by the high levelsof R2 and the low level of errors in calibration and validation. In addi-tion, analysis of fatty acid compositions can be used as complementa-ry data for detecting the presence of RBO in EVOO.

Acknowledgement

Abdul Rohman thanks The Directorate for Higher Education(Dikti), The Ministry of National Education, Republic of Indonesia,for the scholarship during his Ph.D. program in Halal Products Re-search Institute, UPM, Malaysia.

References

[1] I.S. Arvanitoyannis, A. Vlachos, Implementation of physicochemical and sensoryanalysis in conjunction with multivariate analysis towards assessing olive oilauthentication/adulteration, Crit. Rev. Food Sci. Nutr. 47 (2007) 441–498.

[2] R.G. Brereton, Introduction to multivariate calibration in analytical chemistry, An-alyst 125 (2000) 2125–2154.

[3] N. Benoudjit, E. Cools, M. Meurens, M. Verleysen, Chemometric calibration of in-frared spectrometers: selection and validation of variables by non-linear models,Chem. Intel. Lab. Systems 70 (2004) 47–53.

[4] L.E. Rodriguez-Saona, M.E. Allendorf, Use of FTIR for rapid authentication and de-tection of adulteration of food, Annu. Rev. Food Sci. Technol. 2 (2011) 17.1–17.17.

[5] J. Moros, S. Garrigues, M. de la Guardia, Vibrational spectroscopy provides a greentool for multi-component analysis, Trends Anal. Chem. 29 (2009) 578–590.

[6] M. De Luca, W. Terouzi, G. Loele, F. Kzaiber, A. Oussama, F. Oliverio, R. Tauler, G.Ragno, Derivative FTIR spectroscopy for cluster analysis and classification of mo-rocco olive oils, Food Chem. 124 (2011) 1113–1118.

[7] D.L. Garcia-Gonzalez, R. Aparacio, Research in olive oil: challenges for the near fu-ture, J. Agric. Food Chem. 58 (2010) 12569–12577.

[8] T. Gallardo-Velázquez, G. Osorio-Revilla, M. Zuñiga-de Loa, Y. Rivera-Espinoza,Application of FTIR-HATR spectroscopy and multivariate analysis to the quantifi-cation of adulterants in Mexican honeys, Food Res. Int. 42 (2009) 313–318.

[9] A.A. Christy, S. Kasemsumran, Y. Du, K. Ozaki, The detection and quantification ofadulteration in olive oil by near-infrared spectroscopy and chemometrics, Anal.Sci. 20 (2004) 935–940.

[10] M. Arlorio, D.J. Coisson, M. Bordiga, F. Travaglia, C. Garino, L. Zuidmeer, R. Van Ree,M.G. Giuffrida, A. Conti, A. Martelli, Olive oil adulterated with hazelnut oils: sim-ulation to identify possible risks to allergic consumers, Food Addit. Contam. 27(2010) 11–18.

[11] N. Groselj, M. Marjan Vracko, J.A. Fernández Pierna, V. Baeten, M. Novic, The useof FT-MIR spectroscopy and counter-propagation artificial neural networks fortracing the adulteration of olive oil, Acta Chim. Slovenia 55 (2008) 935–941.

[12] D. Özdemir, B. Öztürk, Near infrared spectroscopic determination of olive oiladulteration with sunflower and corn oil, J. Food Drug Anal. 15 (2007) 40–47.

[13] M.J. Lerma-García, G. Ramis-Ramos, J.M. Herrero-Martínez, E.F. Simó-Alfonso, Au-thentication of extra virgin olive oils by Fourier-transform infrared spectroscopy,Food Chem. 118 (2010) 78–83.

[14] N. Vlachos, Y. Skopelitis, M. Psaroudaki, V. Konstantinidou, A. Chatzilazarou, E.Tegou, Applications of Fourier transform-infrared spectroscopy to edible oils,Anal. Chim. Acta 573–574 (2006) 459–465.

[15] A. Rohman, Y.B. Che Man, Fourier transform infrared (FTIR) spectroscopy foranalysis of extra virgin olive oil adulterated with palm oil, Food Res. Int. 43(2010) 886–892.

[16] D. Ballabio, R. Todeschini, Multivariate classification for qualitative analysis, in:D.-W. Sun (Ed.), Infrared Spectroscopy for Food Quality: Analysis and Control,Elsevier, New York, 2009, pp. 83–104.

[17] J.N. Miller, J.C. Miller, Statistics and Chemometrics for Analytical chemistry, 5thEd. Pearson Education Limited, Edinburgh Gate Harlow, 2005.

[18] M. De Luca, F. Oliverio, G. loele, G. Ragno, Multivariate calibration techniques ap-plied to derivative spectroscopy data for analysis of pharmaceutical mixtures,Chem. Intel. Lab. Systems 96 (2009) 14–21.

[19] Y.B. Che Man, A. Rohman, T.S.T. Mansor, Differentiation of lard from other edibleoils by means of Fourier transform infrared spectroscopy and chemometrics, J.Am. Oil Chem. Soc. 88 (2011) 187–192.

[20] A. Rohman, Y.B. Che Man, Monitoring of virgin coconut oil (VCO) adulterationwith palm oil using Fourier transform infrared (FTIR) spectroscopy, J. Food Lipids16 (2009) 618–628.

[21] F. Marini, R. Bucci, I. Ginevro, A.L. Magrì, Coupling of IR measurements and mul-tivariate calibration techniques for the determination of enantiomeric excess inpharmaceutical preparations, Chem. Intel. Lab. Systems 97 (2009) 52–63.

[22] D.L. Pavia, G.M. Lampman, G.S. Kriz-jr, J.R. Vyvyan, Introduction to Spectroscopy,4th Ed. Cencage Learning Inc., Australia, 2009.