Abstract
SAGE has a long tradition of publishing accessible texts explaining key concepts in statistics. Building on this tradition, it has recently published Spatial Statistics & Geostatistics, written by Yongwan Chun and Daniel A. Griffith, as part of their series Advances in Geographic Information Science and Technology.
The book discusses a range of topics in spatial statistics and provides computer code, in R, of analyses presented in the text. R is an open source, higher-level computer language intended for data analysis. It is freely available for Windows, Linux, Unix and Apple OS X, and has a large user community that has provided and continues to provide numerous statistical packages for a wide range of applications. The inclusion of code in R allows users to reproduce examples shown in the book and gives them a starting point for own analysis. A few techniques discussed in the book are based on other computer programs such as SAS, WinBUGS and ArcGIS.
The book is divided into 9 chapters. The first is an introduction, which includes a very brief discussion of R. This is followed by 8 chapters, each of which explains particular aspects of spatial analysis and works towards well-defined learning outcomes. The chapters cover topics such as spatial autocorrelation, spatial sampling methods, spatial composition and configuration (techniques to investigate and account for spatial heterogeneity), spatial regression and econometrics, local statistics, analysis of spatial variance and covariance, interpolation and more advanced topics. Throughout the book, analyses are based on data from Puerto Rico; this provides greater insight and has the advantage that later chapters can refer back to results in earlier chapters. In several cases, alternative techniques are discussed to tackle the same spatial problem. Differences in results between alternative techniques are explained, and potential pitfalls are identified.
The book is in my opinion very useful. I particularly like the choice of statistical problems, the focus on one region to explain a series of problems and the availability of R code, which makes it easy for the reader to reproduce the analysis.
However, I also have a number of reservations. First, the book is vague on the intended audience. The intended reader does need an understanding of a variety of statistical methods. Although each concept is introduced by a brief explanation, it is likely too short for those without much background in statistics. A knowledge of spatial statistics is not required. In some cases, very basic concepts are explained (simple manipulations of numbers and vectors in R), while other concepts, such as installing a range of packages in R (which is very much needed since the book relies on using packages that are not part of the standard R installation), are not explained at all. The latter omission leaves the reader to his or her own devices to dig around the R website (although not a bad thing in itself) to find and install the appropriate packages. The inclusion of one paragraph to explain how to install add-on packages (e.g. type after the R prompt install.packages (‘package.name’) with ‘package.name’ the name of the package to be installed) would have been very useful. A similar criticism can be made on the explanation of statistical concepts, as some are introduced without much discussion and others are introduced twice. For example, the adjustment of significance levels for multiple use of significance tests (the authors use the Bonferroni adjustment) is only partially explained in section 4.1.2 but is more fully and better explained in Chapter 6.
A second criticism is that the writing can be inaccessible, particularly in the earlier chapters. I noticed the occasional awkward sentence such as ‘[…] the bootstrap involves constructing a sample sampling distribution with replicate samples by random sampling with replacement from a single selected sample, using a sample of size n […]’. Fortunately, the style of writing is a lot more fluent in later chapters.
A third, minor criticism is that the main focus of the book is on statistical techniques, but that not much is presented in terms of interpretation of the results. For example, in one of the earlier chapters, a change is noted in the number of farms per area in some regions in Puerto Rico, but no further explanation is provided as to why this may have occurred.
All in all, the book covers a lot of ground: it demonstrates spatial statistics with nice examples in a logical progression and it provides R code for all examples allowing the reader to repeat the analyses in the book. This makes it a valuable book and provides ample compensation for some of the shortcomings.