Practical Three-Way Calibration

Practical Three-Way Calibration

Alejandro C. Olivieri

Graciela M. Escandar

Rosario Institute of Chemistry (IQUIR-CONICET), Department of Analytical Chemistry

Faculty of Biochemical and Pharmaceutical Sciences, National University of Rosario, Suipacha 531, Rosario (2000) Argentina

Table of Contents

Cover image

Title page

Copyright

Dedication

Preface

Foreword

Acknowledgments

Chapter 1. Calibration Scenarios

1.1 Calibration

1.2 Univariate calibration

1.3 Multivariate calibration

1.4 Nomenclature for data and calibrations

1.5 Nomenclature for constituents and samples

1.6 Multiway calibration

1.7 Why multiway calibration?

1.8 Analytical advantages

Chapter 2. Data Properties

2.1 Data properties

2.2 Bilinear data

2.3 Normalization and concentration effects

2.4 A word of caution on bilinearity

2.5 Nonbilinear data

2.6 Trilinear data

2.7 Nontrilinear data

2.8 Transforming three-way data into matrix data

2.9 Normalization and concentration effects

2.10 Classification of three-way data

2.11 Importance of classifying three-way data

Chapter 3. Experimental Three-way/Second-order Data

3.1 Generation of three-way data

3.2 Matrix fluorescence spectroscopy

3.3 Chromatography with spectral detection

Other second-order instrumental data

3.5 Data organization in files

3.6 Samples for calibration and validation

Chapter 4. The MVC2 Software

4.1 Methods, models, algorithms and software

4.2 The MVC2 software

4.3 The MVC2 data examples

4.4 The EEFM_data example

4.5 Plotting EEFM_data matrices

4.6 The LCDAD_data example

4.7 Plotting LCDAD_data matrices

4.8 Further MVC2 features

Chapter 5. Parallel Factor Analysis: Trilinear Data

5.1 Trilinear modeling and decomposition

5.2 Uniqueness and the second-order advantage

5.3 Processing the EEFM_data example

5.4 PARAFAC analysis of a test sample

5.5 Estimating the number of components

5.6 Analyte quantitation in the test sample

5.7 Analysis of the remaining samples

5.8 Profiles for potential interferents

5.9 Further processing options

5.10 Multiple-sample processing

5.11 Concluding remarks

5.12 Homework 1

5.13 Homework 2

Chapter 6. Analytical Figures of Merit

6.1 Definition of figure of merit

6.2 Importance of analytical figures of merit

Sensitivity

6.4 Selectivity

6.5 Analytical sensitivity

6.6 Prediction uncertainty

6.7 Limit of detection

6.8 Limit of quantitation

6.9 The complete PARAFAC report

6.10 Final considerations

Chapter 7. Parallel Factor Analysis: Nontrilinear Data of Type 1

7.1 An apparent contradiction

7.2 Description of the data set

7.3 PARAFAC study of a test sample

7.4 Increasing the number of PARAFAC components

7.5 Study of the remaining samples

7.6 Other separation data and what to do

7.7 A PARAFAC variant for chromatographic data

7.8 PARAFAC2 calibration with the LCDAD_data

7.9 Chromatographic alignment

7.10 Homework

Chapter 8. Multivariate Curve Resolution–Alternating Least-Squares

8.1 Multivariate curve resolution–alternating least-squares

8.2 Estimating the number of components

8.3 MCR–ALS initialization

8.4 Constraints

MCR–ALS analysis of the LCDAD_data set

8.6 Analyte prediction in the test samples

8.7 Analyte prediction in all test samples simultaneously

8.8 Analytical figures of merit

8.9 Conclusion

8.10 Homework 1

8.11 Homework 2

Chapter 9. Partial Least-Squares with Residual Bilinearization

9.1 Introduction

9.2 Unfolded partial least-squares

9.3 Estimating the number of calibration components

9.4 Residual bilinearization

9.5 The EEFM_data set: cross-validation

9.6 The EEFM_data set: RBL and prediction

9.7 Analytical figures of merit

9.8 The LCDAD_data set

9.9 The EEFM_IF_data set

9.10 U-PLS calibration in the EEFM_IF_data set

9.11 RBL in the EEFM_IF_data set

9.12 Other RBL methodologies

9.13 Other Nontrilinear Type 2 data

9.14 The Cinderella type 3 data

9.15 Conclusion

9.16 Homework 1

9.17 Homework 2

9.18 Homework 3

Chapter 10. Three-way/Second-order Standard Addition

10.1 Why standard addition?

10.2 The EEFM_SA example

10.3 Processing the EEFM_SA data set with PARAFAC

10.4 Processing the EEFM_SA data set with MCR–ALS

10.5 Can the EEFM_SA data set be processed with U-PLS/RBL?

Chapter 11. Third-order/Four-way Calibration and Beyond

11.1 Third-order/four-way data

11.2 Generation of third-order/four-way data

Classification of third-order/four-way data

11.4 Algorithms

11.5 Data points in each mode

11.6 Fourth-order/five-way data

11.7 Figures of merit

11.8 Further higher-order advantages

Chapter 12. Application Example: PARAFAC

12.1 Trilinear data

12.2 What algorithm should be chosen?

12.3 A literature EEFM example

12.4 How was the whole experiment designed?

12.5 How were the calibration concentrations chosen?

12.6 How were the validation concentrations chosen?

12.7 How were the wavelength ranges chosen?

12.8 PARAFAC processing using MVC2: validation samples

12.9 What happens for a smaller number of components?

12.10 What happens for a larger number of components?

12.11 Analyte prediction

12.12 Why analyzing test samples?

12.13 PARAFAC processing using MVC2: test samples

12.14 Why analyzing real samples?

Chapter 13. Application Example: MCR–ALS

13.1 Nontrilinear data of Type 1

13.2 How to solve this problem?

13.3 Why coupling multivariate calibration to a separative method?

13.4 A literature example

13.5 Which are the difficulties of aligning chromatographic bands in this complex system?

13.6 What algorithm should be chosen?

13.7 How was the whole experiment designed?

13.8 Preparing the calibration and validation sets

13.9 How were the validation concentrations chosen?

13.10 How were the measuring ranges selected?

13.11 MCR–ALS processing using MVC2: validation samples

13.12 Why test samples should be analyzed?

13.13 MCR–ALS processing using MVC2: test samples

13.14 Analysis of real samples

13.15 Conclusion

Chapter 14. Application Example: U–PLS/RBL

14.1 Nontrilinear data of Type 2

14.2 How to solve this problem?

14.3 What algorithm should be chosen?

14.4 An experimental literature example

14.5 How was the whole experiment designed?

14.6 Preparing the calibration and validation sets

14.7 How to choose the wavelength ranges?

14.8 U-PLS processing using MVC2: validation samples

14.9 Analysis of samples with potential interferences

14.10 Analysis of real samples

14.11 Conclusion

Chapter 15. Solutions to Homework

Homework to Chapter 5

Homework to Chapter 7

Homework to Chapter 8

Homework to Chapter 9

15.5 Conclusion

Index

Copyright

Elsevier

225 Wyman Street, Waltham, MA 02451, USA

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK

Raderweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands

Copyright © 2014 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher

Permissions may be sought directly from Elsevier's Science & Technology Rights Department in Oxford, UK: phone: (+44) 1865 843830, fax: (+44) 1865 853333, E-mail: permissions@elsevier.com. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material

Notice

No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein

Library of Congress Cataloging-in-Publication Data

Application submitted

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-410408-2

For information on all Elsevier publications visit our web site at store.elsevier.com

Printed and bound in USA

14 15 16 17 18 10 9 8 7 6 5 4 3 2 1

Dedication

Dedicated to our beloved daughters and sons Javier, Marcelo, Daniela, Andrea and Sonia, and especially to Raúl, a true booklover, who always gave us the right advice based on his outstanding knowledge.

Preface

God save the readers from long prefaces.¹

This book is about chemometrics. In general terms, chemometrics has the purpose of extracting information from chemical systems by a suitable combination of mathematics, statistics and computer science. It is usually applied to solve chemical problems which can be primarily classified as either descriptive or predictive. In descriptive applications, properties of chemical systems are modeled with the aim of learning underlying relationships and structure, in most cases with a classification scheme in mind, i.e., discriminating objects as belonging to a given class. In predictive applications, on the other hand, the emphasis is on modeling the properties of chemical systems, with the purpose of predicting them in new, unknown specimens. Analytical calibration belongs to this latter type of problems, where the property of interest is typically the concentration of one or more selected analytes in a sample.

This book is also about multiway calibration, and specifically about three-way calibration, which is by far the most pursued sub-division of the latter. It is a relatively new and rapidly growing scientific field, and integrates knowledge from analytical chemistry and chemometrics. An important part of this knowledge is experimental, since the well-known and classical concepts of analytical calibration are preserved in multiway calibration, yet a bonus is added by combining these heavily experimental activities with modeling, regression and statistics, in order to maximize the extraction of information from the raw data and to improve the quality and figures of merit of the analyte predictions.

Based on lectures and courses for graduate chemistry students that we have given in the past years, it was becoming apparent that literature material focusing on practical aspects of multiway calibration rather than on basic chemometrics was required. This book is intended primarily to analytical chemists, and tries to provide a link between the experimental analytical tradition, which is mainly based on classical calibration procedures for single chemical components, and the seemingly more complex calibration scenarios which are today possible thanks to the power of modern instrumentation.

Multiway calibration without formulas is most probably a myth. We have made the effort of leaving a minimum amount of equations, and tried to explain the concepts with words rather than with algebraic expressions. It is likely that some readers will complain that there is still too much math, others may think differently. Overall, we have tried to honor the book title, so the message is basically directed toward chemists who want to appreciate the potentiality of the methods to be described along the book, rather than on chemists trying to deeply understand advanced chemometric concepts. The latter readers will surely find answers to their hopes in other specialized books.²,³

As we are primarily analytical chemists who have learned some chemometrics, we hope the language is understandable for our analytical chemistry colleagues.

References

1. Translated from. De Quevedo Villegas F. El mundo por de dentro. Las Palmas de Gran Canaria, Spain: Red Ediciones; 2011: p. 10.

2. Smilde A., Bro R., Geladi P. Multi-way analysis: applications in the chemical sciences. Chichester: Wiley; 2004.

3. Kroonenberg P.M. Applied multiway data analysis. Chichester: Wiley; 2008.

Foreword

The contents of this book, PRACTICAL THREE-WAY CALIBRATION, by A. C. Olivieri and G. M. Escandar, cover both fundamentals and practice on a subject, multiway calibration, of paramount importance to resolve actual analytical problems in very diverse fields with increasingly complex analytical problems, as environmental, biochemical or food analysis, to name a few areas of interest.

The authors divided the book in two parts, the first one included the Chapters 1 to 4, in which the fundamentals (available techniques, methods and algorithms) and software related to three-way calibration are covered, introducing a basic nomenclature for the different kind of data, describing the data properties and the experimental generation of three-way data, and discussing the advantages of going from classical univariate calibration to multiway calibration.

In a second part, Chapters 5 to 15, several practical applications, including parallel factor analysis (PARAFAC), multivariate curve resolution-alternating least-squares (MCR-ALS) and partial least-squares in combination with residual bilinearization (PLS-RBL), are explained in detail, in order to obtain the maximum possible information from the currently available instrumental data, including the estimation of the analytical figures of merit.

The book is written with analytical chemists in mind, both researches and practitioners, avoiding as much as possible the mathematics involved in these powerful multivariate chemometric tools and, along the text, the advantages of using each of the available methods for each kind of data are highlighted.

The examples chosen, with focus on the most employed data, i.e., excitation-emission fluorescence matrices and chromatographic-spectral matrices, in most instances have been developed by the research group of the authors and collaborators, which is very convenient as the authors have the necessary practical laboratory experience to deal with it.

The necessary use of the suggested free software graphical interfaces, namely the MVC2 and MVC3 Mat Lab routines, to follow the explanations given in the chapters dealing with the applications, opens a very interesting scenario, as the user should be able, after the reading of this book, of exploiting the different chemometric tools available to perform three-way and higher-way calibration with its own laboratory data. Particularly useful in this context is the last chapter with solutions to the homeworks suggested along the practical applications chapters.

In this way, the book is filling a gap in the subject, with an intelligent coverage of the advantages and practical limitations of three-way calibration, by a combination of both fundamentals and practice, and providing free software, developed by the authors, to use the most popular three-way and higher-way available approaches to deal with increasingly complex analytical data.

Arsenio Muñoz de la Peña, Department of Analytical Chemistry, University of Extremadura, Badajoz, Spain

Acknowledgments

Prof. Arsenio Muñoz de la Peña, from University of Extremadura, Spain, for his friendship and continuous support during the last 18 years, and for introducing us to the fascinating world of chemometrics.

Prof. Héctor Goicoechea, University of El Litoral, Argentina, a friend with whom we have enjoyed learning and working in chemometrics.

Prof. Rasmus Bro, University of Copenhagen, Denmark, for making available to the public the highly useful N-way toolbox codes, which are part of the software here described.

Dr. Romà Tauler, Department of Environmental Chemistry (IDAEA-CSIC), Spain, for making available to the public a series of MCR-ALS codes, which inspired our own program codes for the software described in this book.

Profs. Hai-Long Wu and Ru-Qin Yu, Hunan University, China, for generously supplying the program codes of the well-appreciated trilinear decomposition methods, which have been incorporated into the software described in this book.

Facultad de Ciencias Bioquímicas y Farmacéuticas, Universidad Nacional de Rosario, Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) and Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT), for financially supporting our research projects devoted to the main subject of this book.

Our friends, colleagues and research fellows of the Department of Analytical Chemistry and of the Rosario Institute of Chemistry, with whom we shared many years of work and scientific dreams.

Chapter 1

Calibration Scenarios

Abstract

An introduction is provided to various calibration approaches: univariate, multivariate, and multiway, based on the complexity of the collected instrumental data. A basic nomenclature for the different data is introduced, using the complementary concepts of way and order. The advantages in going from classical univariate calibration to multiway calibration methodologies are discussed.

Keywords

Multivariate and multiway data; Multiway calibration advantages; Univariate

Do it your way, but do it multiway!

1.1. Calibration

According to the International Union of Pure and Applied Chemistry (IUPAC), calibration is, in a general sense, an operation that relates an output quantity to an input quantity for a measuring system under given conditions.¹ The input quantities of our primary interest, i.e., in analytical calibration, are the concentrations of a sample constituent of interest (the analyte), while the output quantities are analytical signals or responses delivered by analytical instruments (a spectrometer, chromatograph, voltammeter, etc.). Therefore, in the present context calibration will mean the operation of relating instrumental signals to analyte concentrations.

1.2. Univariate calibration

A specific case of the general calibration process is the one relating the content of a single analyte in a sample to a single value of an instrumental signal, and is called univariate calibration. In analytical chemistry, univariate calibration employs a calibration curve as a general method for the determination of the concentration of a constituent in an unknown sample.¹ The calibration curve is a plot of the instrumental response as a function of the concentration of the analyte. Usually, the analyst prepares a series of standards across a range of concentrations near the expected concentration of the analyte in the unknown sample. The concentrations of the standards lie within the working range of the employed technique (Figure 1.1). Analyzing each of these standards using the chosen technique will produce a series of measurements. For most analyses, a plot of instrumental response vs analyte concentration will show a linear relationship. After measuring the response for the unknown sample, interpolation in the calibration curve allows one to find the concentration of the analyte (Figure 1.1).

1.3. Multivariate calibration

A more general calibration process involves the relationship between the concentrations of various constituents in a test sample and multiple measured responses, i.e., multivariate instead of univariate. In contrast to univariate calibration, which works with a single instrumental response measured for each experimental sample, multivariate calibration works with many different signals for each sample.

Depending on the instrumental setup, the delivered data for a single sample may have different degrees of complexity. The simplest multivariate data are those produced in vector form, i.e., as a series of responses, which can be placed one on top of each other to generate a mathematical object known as a column vector. This object is also referred as having a single mode or direction. Spectra and chromatograms are prime examples of vector data.

FIGURE 1.1   (A) Univariate calibration curve showing a linear relationship between instrumental response (vertical axis) and analyte concentration (horizontal axis). Calibration standards are shown as filled circles, with the solid line implying the best fit to the calibration data. (B) Analyte prediction in an unknown sample: the dashed arrows show the process, as the measured response (open square) is interpolated in the calibration curve, leading to the estimated analyte concentration

Multivariate calibration using vectorial spectral data, particularly those measured through sample absorption in the near infrared (NIR) range, has given rise to a highly fruitful analytical field, which today is routine in many industrial laboratories and process control units.² The literature on this field is abundant. Multiple analytes can be determined simultaneously in the presence of other, possibly unknown constituents, provided they have been properly taken into account during the calibration phase.³,⁴ The latter statement means that for determining an analyte in complex specimens (foodstuff, industrial materials, biological fluids, etc.) a relatively large number of calibration samples is required, in order to be able to model the effect of the interferents in new samples. The field has undoubtedly been triggered by the useful properties of NIR instruments, i.e., the ability to probe samples in a noninvasive manner.²

More complex data arrays than vectors are possible. The increasing complexity derives from increasingly complex instrumental setups, which generate data that can be arranged into mathematical objects with additional modes, such as a matrix or a three-dimensional array (Figure 1.2). In general, all calibration procedures developed using multiple data measured for each sample are known as multivariate. A proper classification of these calibrations depends on how the instrumental data are classified, as detailed in the next section.

FIGURE 1.2   Different Objects, Which can be Built with Data Measured for a Single Sample, Indicating Their Different Modes

1.4. Nomenclature for data and calibrations

Figure 1.2 shows data objects of growing complexity. In algebraic jargon, a scalar is a zeroth-order object, a vector is first order, a matrix is second order, etc. A nomenclature exists for the different calibrations, based on the measurement of data of various orders for a single sample: zeroth-order calibration is equivalent to univariate calibration, first-order multivariate calibration is equivalent to calibration with vectorial data per sample, second-order multivariate calibration is equivalent to calibration with matrix data per sample, and third-order multivariate calibration is equivalent to calibration with three-dimensional data arrays per sample. The list may continue with data arrays with additional modes per sample.⁵

On the other hand, with data for a group of samples it is possible to create an array having an additional mode, the sample mode. For example, univariate measurements for several samples can be grouped to form a vector, first-order data can be placed adjacent to each other to create a matrix, etc. On this basis, an alternative nomenclature has been developed, in which calibrations are named according to the number of ways (modes) of an array for a sample set. Thus zeroth-order calibration is also one-way calibration, first order is two-way, second order is three-way, etc., and it is customary to name all calibration methodologies involving second- and higher-order data (i.e., three-way and beyond) as multiway calibration, which is thus a subdivision of multivariate calibration.

Figure 1.3 provides a compact view of the hierarchy of data and calibrations. Throughout this book, we will employ both the order-based and way-based nomenclatures for calibration.

1.5. Nomenclature for constituents and samples

In order to fully understand the properties of multiway calibration, it is important to define different sample and constituent categories, with particular focus on those generating signals, which may overlap with that of the analyte of interest and can therefore be considered as potential interferents.

The presence of a specific constituent depends on the type of sample being analyzed. Samples can be divided in: (1) calibration or training, (2) validation, (3) test, and (4) real. Calibration samples are employed for establishing the relationship between the known analyte concentrations and the collected instrumental signals. For validation samples, regularly employed to check the analytical performance of the calibration under controlled conditions, the analyte concentration values are also available. These latter samples should have a chemical composition qualitatively similar to the calibration samples, although the analyte concentrations should be different than those employed for calibration, and preferably given by randomly selected values within the corresponding calibration range for each of them. Test samples may contain additional constituents to those present in calibration and validation, and are used in multiway calibration to probe its ability to distinguish the presence of these foreign compounds. Finally, it is always desirable to have a set of real samples, whose composition is expected to be similar to the test samples, and for which analyte concentrations are unavailable, but can be estimated by a reference analytical technique. The real set helps to assess whether the accuracy and precision of the multiway calibration are comparable to those for the reference technique.

FIGURE 1.3   Hierarchy of Data Illustrating the Nomenclatures Based on the Concept of Order and Ways. Top: data for a single sample. Bottom: data for a set of samples. The hypercube corresponding to four-way data is from http://upload.wikimedia.org/wikipedia/commons/5/55/Tesseract.gif, and has been released into the public domain by its author, Jason Hise, at the English Wikipedia project

A distinction can thus be made between constituents present in calibration and validation, and those, which are only present in test and real samples. The former ones can be called expected, because the analyst should include in the calibration set the constituents expected to be present in future samples. The expected constituents could in principle be further divided into calibrated and uncalibrated: calibrated refers to constituents for which calibration concentrations are available, whereas uncalibrated refers to constituents for which only the instrumental signals are measured during the calibration phase, but for which no concentration information is available. We will not employ the latter distinction in this book, mainly because uncalibrated might be easily confused for unexpected.

Other constituents present in test and real samples, beyond those in the calibration set, are called unexpected, and always carry the potential of interfering in the analysis. However, potential interferents may not always be harmful. The official definition of interference is anything generating a systematic error in the analyte determination.⁶ In general, whether the interference will be actual or will only remain as potential, depends on the type of measured instrumental signals and on the employed calibration methodology. For univariate and first-order instrumental data, unexpected, unknown sample constituents are likely to interfere. This may not be true in multiway calibration, as we shall repeatedly see along this book.

Table 1.1 summarizes the various sample and constituent types.

Table 1.1

Presence of the Various Constituent Types in Different Kind of Samples

1.6. Multiway calibration

Multiway calibration is based on many instrumental signals per sample, which can be meaningfully organized into a certain mathematical object with more modes than a vector, for example, as a data table or matrix.⁷ Moreover, the data can be organized in such an object because of a strong underlying reason for doing so. For example, a single UV–visible absorption spectrum measured at 625 wavelengths can in principle be transformed into a data matrix of size 25 × 25, or into a data matrix of size 125 × 5; however, why would one do such transformations? There is no clear rationale supporting them. In contrast, using a modern spectrofluorimeter it is possible to measure 25 emission spectra at 25 different excitation wavelengths, producing a 25 × 25 data matrix, which can be logically saved as a data table having in each column an emission spectrum obtained at a certain excitation wavelength, and in each row an excitation spectrum obtained at a certain emission wavelength. This data table could be transformed from matrix to vector format, by concatenating or unfolding the matrix column by column into a 625 × 1 one-column vector. This will destroy, so to speak, the logical organization of the measured data, leading to loss of information.

The above mentioned fluorescence matrix data can be graphically visualized in several different representations (Figure 1.4): (1) as a series of lines connecting the values contained in the 25 columns of the data table, each line representing an emission spectrum, (2) as a series of lines connecting the values contained in the 25 rows, each line representing an excitation spectrum, (3) as a mesh plot, connecting both columns and rows, and (4) as a contour plot showing successive curves obtained by projection of the surface onto the excitation–emission wavelength plane. The mesh plot is also popularly known as a landscape, and it is not uncommon to talk about fluorescence landscapes. The three axes of this landscape are fluorescence intensity, excitation wavelength, and emission wavelength.

Another pertinent example of a data matrix is the one produced by a liquid chromatograph (LC) connected to a UV–visible diode array detector (called LC-DAD data). At certain time intervals, the detector measures spectra in a given wavelength range. These data can be logically saved as a table having in each column a chromatogram obtained at a certain wavelength, and in each row a spectrum measured at a certain elution time. LC-DAD data will produce a landscape similar to that of Figure 1.4, whose axes are absorbance, elution time, and wavelength.

In principle, one can foresee no limit as to the number of different data modes, which could be experimentally accessible with future instrumentation. In practice, the most usual setup is to measure data matrices, with still few works collecting three-dimensional arrays per sample.

FIGURE 1.4   Different Graphical Representations of an Excitation–Emission Fluorescence Data Matrix. (A) Emission spectra as a function of excitation wavelength (columns of the data matrix). (B) Excitation spectra as a function of emission wavelength (rows of the data matrix). (C) Mesh plot connecting both columns and rows. (D) Contour plot

Other than that, the