*551*
*72*
*4MB*

*English*
*Pages XII, 316 s. : il. kolor. ; 24 cm
[329]*
*Year 2016*

Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan

This page intentionally left blank

Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan

Fra¨nzi Korner-Nievergelt Tobias Roth Stefanie von Felten Je´roˆme Gue´lat Bettina Almasi Pius Korner-Nievergelt

AMSTERDAM l BOSTON l HEIDELBERG l LONDON NEW YORK l OXFORD l PARIS l SAN DIEGO SAN FRANCISCO l SINGAPORE l SYDNEY l TOKYO Academic Press is an imprint of Elsevier

Academic Press is an imprint of Elsevier 32 Jamestown Road, London NW1 7BY, UK 525 B Street, Suite 1800, San Diego, CA 92101-4495, USA 225 Wyman Street, Waltham, MA 02451, USA The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK Copyright Ó 2015 Elsevier Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangement with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability 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. ISBN: 978-0-12-801370-0 British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress For information on all Academic Press Publications visit our website at http://store.elsevier.com/ Printed and bound in the USA

Contents Digital Assets Acknowledgments

1.

Why do we Need Statistical Models and What is this Book About? 1.1 1.2

2.

1 1 2 4

Prerequisites and Vocabulary

5

2.1

5 5 6

2.2

3.

Why We Need Statistical Models What This Book is About Further Reading

x xi

Software 2.1.1 What Is R? 2.1.2 Working with R Important Statistical Terms and How to Handle Them in R 2.2.1 Data Sets, Variables, and Observations 2.2.2 Distributions and Summary Statistics 2.2.3 More on R Objects 2.2.4 R Functions for Graphics 2.2.5 Writing Our Own R Functions Further Reading

7 7 12 15 16 17 18

The Bayesian and the Frequentist Ways of Analyzing Data

19

3.1 3.2

19 20

3.3 3.4

Short Historical Overview The Bayesian Way 3.2.1 Estimating the Mean of a Normal Distribution with a Known Variance 3.2.2 Estimating Mean and Variance of a Normal Distribution Using Simulation The Frequentist Way Comparison of the Bayesian and the Frequentist Ways Further Reading

21 22 27 30 31

v

vi Contents

4.

Normal Linear Models

33

4.1

33 33 36 36 41

4.2

5.

6.

7.

Linear Regression 4.1.1 Background 4.1.2 Fitting a Linear Regression in R 4.1.3 Drawing Conclusions 4.1.4 Frequentist Results Regression Variants: ANOVA, ANCOVA, and Multiple Regression 4.2.1 One-Way ANOVA 4.2.2 Frequentist Results from a One-Way ANOVA 4.2.3 Two-Way ANOVA 4.2.4 Frequentist Results from a Two-Way ANOVA 4.2.5 Multiple Comparisons and Post Hoc Tests 4.2.6 Analysis of Covariance 4.2.7 Multiple Regression and Collinearity 4.2.8 Ordered Factors and Contrasts 4.2.9 Quadratic and Higher Polynomial Terms Further Reading

42 42 47 49 53 55 56 60 64 66 67

Likelihood

69

5.1 5.2 5.3

69 72 73 74

Theory The Maximum Likelihood Method The Log Pointwise Predictive Density Further Reading

Assessing Model Assumptions

75

6.1 6.2 6.3 6.4 6.5 6.6

75 76 79 80 85 91 94

Model Assumptions Independent and Identically Distributed The QQ Plot Temporal Autocorrelation Spatial Autocorrelation Heteroscedasticity Further Reading

Linear Mixed Effects Models 7.1

7.2 7.3 7.4 7.5 7.6 7.7

Background 7.1.1 Why Mixed Effects Models? 7.1.2 Random Factors and Partial Pooling Fitting a Linear Mixed Model in R Restricted Maximum Likelihood Estimation Assessing Model Assumptions Drawing Conclusions Frequentist Results Random Intercept and Random Slope

95 95 95 96 98 101 102 103 105 106

Contents vii

7.8 7.9

8.

115

8.1 8.2

115 117 117 118

8.4

Background Binomial Model 8.2.1 Background 8.2.2 Fitting a Binomial Model in R 8.2.3 Assessing Model Assumptions: Overdispersion and Zero-Inflation 8.2.4 Drawing Conclusions 8.2.5 Frequentist Results Fitting a Binary Logistic Regression in R 8.3.1 Some Final Remarks Poisson Model 8.4.1 Background 8.4.2 Fitting a Poisson-Model in R 8.4.3 Assessing Model Assumptions 8.4.4 Drawing Conclusions 8.4.5 Modeling Rates and Densities: Poisson Model with an Offset 8.4.6 Frequentist Results Further Reading

121 126 127 128 131 132 132 134 135 136 137 138 139

Generalized Linear Mixed Models

141

9.1

141 141 142 143 145 146 146 146 148 149

9.2

10.

111 114 114

Generalized Linear Models

8.3

9.

Nested and Crossed Random Effects Model Selection in Mixed Models Further Reading

Binomial Mixed Model 9.1.1 Background 9.1.2 Fitting a Binomial Mixed Model in R 9.1.3 Assessing Model Assumptions 9.1.4 Drawing Conclusions Poisson Mixed Model 9.2.1 Background 9.2.2 Fitting a Poisson Mixed Model in R 9.2.3 Assessing Model Assumptions 9.2.4 Drawing Conclusions 9.2.5 Modeling Bird Densities by a Poisson Mixed Model Including an Offset Further Reading

151 159

Posterior Predictive Model Checking and Proportion of Explained Variance 161 10.1 10.2

Posterior Predictive Model Checking Measures of Explained Variance Further Reading

161 169 174

viii Contents

11.

Model Selection and Multimodel Inference 11.1 11.2

11.3 11.4

12.

14.

175 179 179 181 183 184 184 185 188 193 196

Markov Chain Monte Carlo Simulation

197

12.1 12.2

197 201 202 204 206 211 212

12.3 12.4

13.

When and Why We Select Models and Why This is Difficult Methods for Model Selection and Model Comparisons 11.2.1 Cross-Validation 11.2.2 Information Criteria: Akaike Information Criterion and Widely Applicable Information Criterion 11.2.3 Other Information Criteria 11.2.4 Bayes Factors and Posterior Model Probabilities 11.2.5 Model-Based Methods to Obtain Posterior Model Probabilities and Inclusion Probabilities 11.2.6 “Least Absolute Shrinkage and Selection Operator” (LASSO) and Ridge Regression Multimodel Inference Which Method to Choose and Which Strategy to Follow Further Reading

175

Background MCMC Using BUGS 12.2.1 Using BUGS from OpenBUGS 12.2.2 Using BUGS from R MCMC Using Stan Sim, BUGS, and Stan Further Reading

Modeling Spatial Data Using GLMM

213

13.1 13.2 13.3

213 214 214 214 216 216 222 224

Background Modeling Assumptions Explicit Modeling of Spatial Autocorrelation 13.3.1 Starting the Model Fitting 13.3.2 Variogram Modeling 13.3.3 Bayesian Modeling 13.3.4 OpenBUGS Example Further Reading

Advanced Ecological Models 14.1 14.2

Hierarchical Multinomial Model to Analyze Habitat Selection Using BUGS Zero-Inflated Poisson Mixed Model for Analyzing Breeding Success Using Stan

225 225 231

Contents

14.3 14.4 14.5

15.

16.

17.

240 246 252 263

Prior Influence and Parameter Estimability

265

15.1 15.2 15.3

265 269 273 278

How to Specify Prior Distributions Prior Sensitivity Analysis Parameter Estimability Further Reading

Checklist

279

16.1

279 287

Data Analysis Step by Step Further Reading

What Should I Report in a Paper

289

17.1 17.2

289 293 296

References Index

Occupancy Model to Measure Species Distribution Using Stan Territory Occupancy Model to Estimate Survival Using BUGS Analyzing Survival Based on Mark-Recapture Data Using Stan Further Reading

ix

How to Present the Results How to Write Up the Statistical Methods Further Reading

297 309

Digital Assets Thank you for selecting Academic Press’ Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan. To complement the learning experience, we have provided a number of online tools to accompany this edition. To view the R-package “blmeco” that contains all example data and some specific functions presented in the book, visit www.r-project.org. The full R-Code and exercises for each chapter are provided at www.oikostat.ch/blmeco.htm.

x

Acknowledgments The basis of this book is a course script written for statistics classes at the International Max Planck Research School for Organismal Biology (IMPRS)dsee www.orn.mpg.de/2453/Short_portrait. We, therefore, sincerely thank all the IMPRS students who have used the script and worked with us. The text grew as a result of questions and problems that appeared during the application of linear models to the various Ph.D. projects of the IMPRS students. Their enthusiasm in analyzing data and discussion of their problems motivated us to write this book, with the hope that it will be of help to future students. We especially thank Daniel Piechowski and Margrit Hieber-Ruiz for hiring us to give the courses at the IMPRS. The main part of the book was written by FK and PK during time spent at the Colorado Cooperative Fish and Wildlife Research Unit and the Department of Fish, Wildlife, and Conservation Biology at Colorado State University in the spring of 2014. Here, we were kindly hosted and experienced a motivating time. William Kendall made this possible, for which we are very grateful. Gabriele Engeler and Joyce Pratt managed the administrational challenges of tenure there and made us feel at home. Allison Level kindly introduced us to the CSU library system, which we used extensively while writing this book. We enjoyed a very inspiring environment and cordially thank all the Fish, Wildlife, and Conservation Biology staff and students who we met during our stay. The companies and institutions at which the authors were employed during the work on the book always positively supported the project, even when it produced delays in other projects. We are grateful to all our colleagues at the Swiss Ornithological Institute (www.vogelwarte.ch), oikostat GmbH (www.oikostat.ch), Hintermann & Weber AG (www.hintermannweber.ch), the University of Basel, and the Clinical Trial Unit at the University of Basel (www.scto.ch/en/CTU-Network/CTU-Basel.html). We are very grateful to the R Development Core Team (http://www. r-project.org/contributors.html) for providing and maintaining this wonderful software and network tool. We appreciate the flexibility and understandability of the language R and the possibilitiy to easily exchange code. Similarly, we would like to thank the developers of BUGS (http://www.openbugs.net/w/ BugsDev) and Stan (http://mc-stan.org/team.html) for making all their extremely useful software freely available. Coding BUGS or Stan has helped

xi

xii Acknowledgments

us in many cases to think more clearly about the biological processes that have generated our data. Example data were kindly provided by the Ulmet-Kommission (www.bnv. ch), the Landschaft und Gewa¨sser of Kanton Aargau, the Swiss Ornithological Institute (www.vogelwarte.ch), Valentin Amrhein, Anja Bock, Christoph Bu¨hler, Karl-Heinz Clever, Thomas Gottschalk, Martin Gru¨ebler, Gu¨nther Herbert, Thomas Hoffmeister, Rainer Holler, Beat Naef-Daenzer, Werner Peter, Luc Schifferli, Udo Seum, Maris Strazds, and Jean-Luc Zollinger. For comments on the manuscript we thank Martin Bulla, Kim MeichtryStier and Marco Perrig. We also thank Roey Angel, Karin Boos, Paul Conn, and two anonymous reviewers for many valuable suggestions regarding the book’s structure and details in the text. Holger Schielzeth gave valuable comments and input for Chapter 10, and David Anderson and Michael Schaub commented on Chapter 11. Bob Carpenter figured out essential parts of the Stan code for the CormackeJollyeSeber model. Michael Betancourt and Bob Carpenter commented on the introduction to MCMC and the Stan examples. Valentin Amrhein and Barbara Helm provided input for Chapter 17. All these people greatly improved the quality of the book, made the text more accessible, and helped reduce the error rate. Finally, we are extremely thankful for the tremendous work that Kate Huyvaert did proofreading our English.

Chapter 1

Why do we Need Statistical Models and What is this Book About? Chapter Outline

1.1 Why We Need Statistical Models

1.2 What This Book is About

2

1

1.1 WHY WE NEED STATISTICAL MODELS There are at least four main reasons why statistical models are used: (1) models help to describe how we think a system works, (2) data can be summarized using models, (3) comparison of model predictions with data helps with understanding the system, and (4) models allow for predictions, including the quantification of their uncertainty, and, therefore, they help with making decisions. A statistical model is a mathematical construct based on probability theory that aims to reconstruct the system or the process under study; the data are observations of this system or process. When we speak of “models” in this book, we always mean statistical models. Models express what we know (or, better, what we think we know) about a natural system. The difference between the model and the observations shows that what we think about the system may still not be realistic and, therefore, points out what we may want to think about more intensively. In this way, statistical models help with understanding natural systems. Analyzing data using statistical models is rarely just applying one model to the data and extracting the results. Rather, it is an iterative process of fitting a model, comparing the model with the data, gaining insight into the system from specific discrepancies between the model and the data, and then finding a more realistic model. Analyzing data using statistical models is a learning process. Reality is usually too complex to be perfectly represented by a model. Thus, no model is perfect, but a good model is useful (e.g., Box, 1979). Often, several models may be plausible and fit the data reasonably well. In such cases, the inference can be based on the set of all models, or a model that performs best for Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan http://dx.doi.org/10.1016/B978-0-12-801370-0.00001-0. Copyright © 2015 Elsevier Inc. All rights reserved.

1

2 Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan

a specific purpose is selected. In Chapter 11 we have compiled a number of approaches we found useful for model comparisons and multimodel inference. Once we have one or several models, we want to draw inferences from the model(s). Estimates of the effects of the predictor variables on the outcome variables, fitted values, or derived quantities that are of biological interest are extracted, together with an uncertainty estimate. In this book we use, except in one example, Bayesian methods to assess uncertainty of the estimates. Models summarize data. When we have measured the height of 100 trees in a forest and we would like to report these heights to colleagues, we report the mean and the standard deviation instead of reporting all 100 values. The mean and the standard deviation, together with a distributional assumption (e.g., the normal distribution) represent a statistical model that describes the data. We do not need to report all 100 values because the 2 values (mean and standard deviation) describe the distribution of the 100 values sufficiently well so that people have a picture of the heights of the 100 trees. With increasing complexity of the data, we need more complex models that summarize the data in a sensible way. Statistical models are widely applied because they allow for quantifying uncertainty and making predictions. A well-known application of statistical models is the weather forecast. Additional examples include the prediction of bird or bat collision risks at wind energy turbines based on some covariates, the avalanche bulletins, or all the models used to predict changes of an ecosystem when temperatures rise. Political decisions are often based on models or model predictions. Models are pervasive; they even govern our daily life. For example, we first expected our children to get home before 3:30 p.m. because we knew that the school bus drops them off at 3:24, and a child can walk 200 m in around 4 min. What we had in mind was a model child. After some weeks observing the time our children came home after school, we could compare the model prediction with real data. Based on this comparison and short interviews with the children, we included “playing with the neighbor’s dog” in our model and updated the expected arrival time to 3:45 p.m.

1.2 WHAT THIS BOOK IS ABOUT This book is about a broad class of statistical models called linear models. Such models have a systematic part and a stochastic part. The systematic part describes how the outcome variable (y, variable of interest) is related to the predictor variables (x, explanatory variables). This part produces the fitted values that are completely defined by the values of the predictor variables. The stochastic part of the model describes the scatter of the observations around the fitted values using a probability distribution. For example, a regression line is the systematic part of the model, and the scatter of the data around the regression line (more precisely: the distribution of the residuals) is the stochastic part.

Why do we Need Statistical Models Chapter j 1

3

Linear models are probably the most commonly used models in biology and in many other research areas. Linear models form the basis for many statistical methods such as survival analysis, structural equation analysis, variance components analysis, time-series analysis, and most multivariate techniques. It is of crucial importance to understand linear models when doing quantitative research in biology, agronomy, social sciences, and so on. This book introduces linear models and describes how to fit linear models in R, BUGS, and Stan. The book is written for scientists (particularly organismal biologists and ecologists; many of our examples come from ecology). The number of mathematical formulae is reduced to what we think is essential to correctly interpret model structure and results. Chapter 2 provides some basic information regarding software used in this book, important statistical terms, and how to work with them using the statistical software package R, which is used in most chapters of the book. The linear relationship between the outcome y and the predictor x can be straightforward, as in linear models with normal error distribution (normal linear model, LM, Chapter 4). But the linear relationship can also be indirect via a link function. In this case, the direct linear relationship is between a transformed outcome variable and the predictor variables, and, usually, the model has a nonnormal error distribution such as Poisson or binomial (generalized linear model, GLM, Chapter 8). Generalized linear models can handle outcome variables that are not on a continuous and infinite scale, such as counts and proportions. For some linear models (LM, GLM) the observations are required to be independent of each other. However, this is often not the case, for example, when more than one measurement is taken on the same individual (i.e., repeated measurements) or when several individuals belong to the same nest, farm, or another grouping factor. Such data should be analyzed using mixed models (LMM, GLMM, Chapters 7 and 9); they account for the nonindependence of the observations. Nonindependence of data may also be introduced when observations are made close to each other (in space or time). In Chapter 6 we show how temporal or spatial autocorrelation is detected and we give a few hints about how temporal correlation can be addressed. In Chapter 13, we analyze spatial data using a species distribution example. Chapter 14 contains examples of more complex analyses of ecological data sets. These models should be understandable with the theory learned in the first part of the book. The chapter presents ideas on how the linear model can be expanded to more complex models. The software BUGS and Stan, introduced in Chapter 12, are used to fit these complex models. BUGS and Stan are relatively easy to use and flexible enough to build many specific models. We hope that this chapter motivates biologists and others to build their own models for the particular process they are investigating. Throughout the book, we treat model checking using graphical methods with high importance. Residual analysis is discussed in Chapter 6. Chapter 10

4 Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan

introduces posterior predictive model checking. Posterior predictive model checking is used in Chapter 14 to explore the performance of more complex models such as a zero-inflated and a territory occupancy model. Finally, in Chapter 15, we present possible ways to assess prior sensitivity. The aim of the checklist in Chapter 16 is to guide scientists through a data analysis. It may be used as a look-up table for choosing a type of model depending on the data type, deciding whether to transform variables or not, deciding which test statistic to use in posterior predictive model checking, or understanding what may help when the residual plots do not look as they should. Such look-up tables cannot be general and complete, but the suggestions they make can help when starting an analysis. For the reasons explained in Chapter 3, we use Bayesian methods to draw inference from the models throughout the book. However, the book is not a thorough introduction to Bayesian data analysis. We introduce the principles of Bayesian data analysis that we think are important for the application of Bayesian methods. We start simply by producing posterior distributions for the model parameters of linear models fitted in the widely used open source software R (R Core Team, 2014). In the second part of the book, we introduce Markov chain Monte Carlo simulations for non-mathematicians and use the software OpenBUGS (Lunn et al., 2013) and Stan (mc-stan.org). The third part of the book includes, in addition to the data analysis checklist, example text for the presentation of results from a Bayesian data analysis in a paper. We also explain how the methods presented in the book can be described in the methods section of a paper. Hopefully, the book provides a gentle introduction to applied Bayesian data analysis and motivates the reader to deepen and expand knowledge about these techniques, and to apply Bayesian methods in their data analyses.

FURTHER READING Gelman and Hill (2007) teach Bayesian data analysis using linear models in a very creative way, with examples from the social and political sciences. Kruschke (2011) gives a thorough and very understandable introduction to Bayesian data analysis. McCarthy (2007) concisely introduces Bayesian methods using WinBUGS. Ke´ry (2010) gives an introduction to linear models using Bayesian methods with WinBUGS. Stauffer (2008) works practically through common research problems in the life sciences using Bayesian methods. Faraway (2005, 2006) and Fox and Weisberg (2011) provide applied introductions to linear models using frequentist methods in R. Note that there is an extensive erratum to Faraway (2006) on the web. Zuur et al. (2009, 2012) are practical and understandable introductions to linear models in R with a particular focus on complex real ecological data problems such as nonindependent data. Zuur et al. (2012) also introduce Bayesian methods. A more theoretical approach, including R code, is Aitkin et al. (2009). We can also recommend the chapters introducing generalized linear models in Wood (2006).

Chapter 2

Prerequisites and Vocabulary Chapter Outline

2.1 Software 2.1.1 What Is R? 2.1.2 Working with R 2.2 Important Statistical Terms and How to Handle Them in R 2.2.1 Data Sets, Variables, and Observations

5 5 6

7

2.2.2 Distributions and Summary Statistics 2.2.3 More on R Objects 2.2.4 R Functions for Graphics 2.2.5 Writing Our Own R Functions

12 15 16 17

7

2.1 SOFTWARE In most chapters of this book we work with the statistical software R (R Core Team, 2014). R is a very powerful tool for statistics and graphics in general. However, it is limited with regard to Bayesian methods applied to more complex models. In Part II of the book (Chapters 12e15), we therefore use Open BUGS (www.openbugs.net; Spiegelhalter et al., 2007) and Stan (Stan Development Team, 2014), using specific interfaces to operate them from within R. OpenBUGS and Stan are introduced in Chapter 12. Here, we briefly introduce R.

2.1.1 What Is R? R is a software environment for statistics and graphics that is free in two ways: free to download and free source code (www.r-project.org). The first version of R was written by Robert Gentleman and Ross Ihaka of the University of Auckland (note that both names begin with “R”). Since 1997, R has been governed by a core group of R contributors (www.r-project.org/contributors. html). R is a descendant of the commercial S language and environment that was developed at Bell Laboratories by John Chambers and colleagues. Most code written for S runs in R, too. It is an asset of R that, along with statistical analyses, well-designed publication-quality graphics can be produced. R runs on all operating systems (UNIX, Linux, Mac, Windows). Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan http://dx.doi.org/10.1016/B978-0-12-801370-0.00002-2. Copyright © 2015 Elsevier Inc. All rights reserved.

5

6 Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and Stan

R is different from many statistical software packages that work with menus. R is a programming language or, in fact, a programming environment. This means that we need to write down our commands in the form of R code. While this may need a bit of effort in the beginning, we will soon be able to reap the first fruits. Writing code enforces us to know what we are doing and why we are doing it, and enables us to learn about statistics and the R language rapidly. And because we save the R code of our analyses, they are easily reproduced, comprehensible for colleagues (especially if the code is furnished with comments), or easily adapted and extended to a similar new analysis. Due to its flexibility, R also allows us to write our own functions and to make them available for other users by sharing R code or, even better, by compiling them in an R package. R packages are extensions of the slim basic R distribution, which is supplied with only about eight packages, and typically contain R functions and sometimes also data sets. A steadily increasing number of packages are available from the network of CRAN mirror sites (currently over 5000), accessible at www.r-project.org. Compared to other dynamic, high-level programming languages such as Python (www.python.org) or Julia (Bezanson et al., 2012; www.julialang.org), R will need more time for complex computations on large data sets. However, the aim of R is to provide an intuitive, “easy to use” programming language for data analyses for those who are not computer specialists (Chambers, 2008), thereby trading off computing power and sometimes also precision of the code. For example, R is quite flexible regarding the use of spaces in the code, which is convenient for the user. In contrast, Python and Julia require a stricter coding, which makes the code more precise but also more difficult to learn. Thus, we consider R as the ideal language for many statistical problems faced by ecologists and many other scientists.

2.1.2 Working with R If you are completely new to R, we recommend that you take an introductory course or work through an introductory book or document (see recommendations in the Further Reading section at the end of this chapter). R is organized around functions, that is, defined commands that typically require inputs (arguments) and return an output. In what follows, we will explain some important R functions used in this book, without providing a full introduction to R. Moreover, the list of functions explained in this chapter is only a selection and we will come across many other functions in this book. That said, what follows should suffice to give you a jumpstart. We can easily install additional packages by using the function install.packages and load packages by using the function library. Each R function has documentation describing what the function does and how it is used. If the package containing a function is loaded in the current R session, we can open the documentation using ?. Typing ?mean into the R

Prerequisites and Vocabulary Chapter j 2

7

console will open the documentation for the function mean (arithmetic mean). If we are looking for a specific function, we can use the function help.search to search for functions within all installed packages. Typing help. search(“linear model”), will open a list of functions dealing with linear models (together with the package containing them). For example, stats::lm suggests the function lm from the package stats. Shorter, but equivalent to help.search(“linear model”) is ??”linear model”. Alternatively, R’s online documentation can also be accessed with help.start(). Functions/packages that are not installed yet can be found using the specific search menu on www. r-project.org. Once familiar with using the R help and searching the internet efficiently for R-related topics, we can independently broaden our knowledge about R. Note that whenever we show R code in this book, the code is printed in orange font. Comments, which are preceded by a hash sign, #, and are therefore not executed by R, are printed in green. R output is printed in blue font.

2.2 IMPORTANT STATISTICAL TERMS AND HOW TO HANDLE THEM IN R 2.2.1 Data Sets, Variables, and Observations Data are always collected on a sample of objects (e.g., animals, plants, or plots). An observation refers to the smallest observational or experimental unit. In fact, this can also be a smaller unit, such as the wing of a bird, a leaf of a plant, or a subplot. Data are collected with regard to certain characteristics (e.g., age, sex, size, weight, level of blood parameters), all of which are called variables. A collection of data, a so-called “data set,” can consist of one or many variables. The term variable illustrates that these characteristics vary between the observations. Variables can be classified in several ways, for instance, by the scale of measurement. We distinguish between nominal, ordinal, and numeric variables (see Table 2-1). Nominal and ordinal variables can be summarized as categorical variables. Numeric variables can be further classified as discrete or continuous. Moreover, note that categorical variables are often called factors and numeric variables are often called covariates. Now let us look at ways to store and handle data in R. A simple, but probably the most important, data structure is a vector. It is a collection of ordered elements of the same type. We can use the function c to combine these elements, which are automatically coerced to a common type. The type of elements determines the type of the vector. Vectors can (among other things) be used to represent variables. Here are some examples: v1