See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/317083600
Multivariate Statistical Approaches in Archeology: A Systematic Review
Article · May 2017
CITATIONS READS
0 895
2 authors:
Abdolkarim Shadmehr Shayan Mostafaei
Tarbiat Modares University Tehran University of Medical Sciences
6 PUBLICATIONS 6 CITATIONS 109 PUBLICATIONS 428 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Gene expression profile of proinflammatory cytokines in Iranian patients with Ankylosing Spondylitis View project
The role of noncoding-RNAs in Cancer and Infectious disease View project
All content following this page was uploaded by Shayan Mostafaei on 23 May 2017.
The user has requested enhancement of the downloaded file.
International Journal of the Society of Iranian Archaeologists Vol. 2, No. 4, Summer- Autumn 2016
Multivariate Statistical Approaches in Archeology:
A Systematic Review
Abdolkarim Shadmehr
Iranian Center for Archaeological Research
Shayan Mostafaei
Tarbiat Modares University
Received: October, 8, 2016 Accepted: December, 12, 2016
Abstract: One of the most important achievements of New Archaeology was a tendency toward application of variables and statistical
analysis in archaeological research. Each statistical method should be applied to certain areas of archaeological studies; it is important
to extract biological, economic, social and cultural information properly via statistical methods. In this study, which is a review through
previously published research, articles in which multivariate statistical method are applied in archaeological investigations have been
extracted from three worldwide databases, i.e. PubMed, Scopus and Science Direct, based on a protocol designed to search for related
articles from January 2000 to January 2016. After application of inclusion and exclusion criteria, finally 384 articles were selected
for this investigation. All of the 384 articles were classified based on multivariate statistical methods and then the application of these
methods in archaeology and cultural material types was determined. They show that methods, including Cluster analysis, Principal
Component Analysis, Discriminant Analysis, Multivariate Multiple Regression, Factor Analysis and Multidimensional Scaling have
had respectively the highest application in archaeological investigations. The results of this systematic review indicate that cluster
analysis is one of the most applied statistical analysis, perhaps because of usage, method and simple interpretation, compared to other
methods of data reduction. This method is used for data reduction or clustering archaeological sites based on their similarities and
helps with the comparison between site structures. Principle Component Analysis is the second most widely used methods due to its
application in any data structure and simplicity of interpretation compared to other methods of dimensionality reduction.
Keywords: Multivariate, Archaeology, Cluster Analysis, Principle Component Analysis.
Introduction
New Archaeology as a new school of archaeological this process many researchers around the world have
studies began in the 1960s as a result of efforts of Lewis used statistical techniques to analyze the archaeological
Binford and David L. Clarke. One of the most important material, for example L. Xiancen and his colleague who
achievements of the new school was a tendency toward studied the use of statistics and mathematics in processing
application of variables and statistical analysis in archaeological data in 1995 (Xiancen and Renping 1995).
archaeological research. Hence archaeology passed the Then Baxter from the department of statistics of Bowling
stage of description and entered into a stage of scientific Green State University used exploratory multivariate
analysis. Because statistics with its unique power to change analysis in archaeology (Baxter 2009). Then J. de Leeuw
the language of data from the description to numbers and from the department of California, Los Angeles university
analyzing the numbers helps with clarifying the situation was the first researcher who introduced Correspondence
of hypotheses in various science fields (Gowland and Analysis in archaeology in 2007 (de Leeuw 2007). From
Western 2012). In fact, application of statistical methods
Abdolkarim Shadmehr
had been introduced in archaeology since 1950s, but it
Deputy of Financial Affairs and Management Development
was only used in descriptive and discovering the structure Iranian Center for Archaeological Research & Faculty Member,
of sites, comparison of sites and distinction of cultural Tehran, Iran.
finds (Myers 1950). Since 1975 statistical methods were k.shadmehr@richt.ir
used in different archaeological studies, for example
Karim Najafi Barzegar
advanced statistical methods were used in geological
corresponding auther,
studies in Egypt, the results of which were published Department of Biostatistics, Faculty of Medical Sciences, Tarbiat
as a book chapter (El Shazly 1957). Then DH Thomas Modares University, Tehran, Iran.
investigated about statistics in archaeology in 1978. By Mostafa.shayan@modares.ac.ir
88
Vol. 2, No. 4, Summer- Autumn 2016 International Journal of the Society of Iranian Archaeologists
2000, application of Multivariate statistical methods articles by Enayatollah Amirlu and Abdolrahman Rasekh
was increased remarkably, according to the published (Rezalo 2009), Kamalaldin Niknami who has published
articles. In 2008, Philip L. Walker from the Department of two educational books for students, statistical methods
Anthropology, university of California used Discriminant are almost ignored by the archaeologists. On the other
Analysis for sex determination in skulls of different hand, there seems to be a lack of systematic reviews
nationalities which had an accuracy of forecasts of over over multivariate statistical methods in archaeological
80% (Walker 2008). Drennan started educating theories and studies. In this article, a systematic review of the articles
applications of statistical methods, especially multivariate representing applications of multivariate statistical methods
methods in archaeology by publishing a valuable book in archaeological studies is provided using the worldwide
(Drennan 2009). The book serves as an educational valid databases including Scopus, PubMed and ScienceDirect.
reference in statistical methods used in archaeology around Various applications of multivariate statistical methods
the world. in different archaeological areas are investigated and
Statistics is a group of strategies to collect the data and examined.
information in order to achieve a certain purpose; it is a
useful means of interpretation of the collected information Materials and Methods
to make reasonable and correct decisions. In other words,
statistics is a group of methods in which logical order and In this study, which is a review through previously published
statistical reasoning is used to analyze data and finally research, articles in which multivariate statistical method is
inferences are made based on the analysis. After several applied in archaeological investigations have been extracted
decades of archaeological excavations, wide databases from three worldwide databases, i.e. PubMed, Scopus and
have been created and developed around the world. Science Direct, based on a protocol designed to search for
Archaeology as a science dealing with massive amounts related articles from January 2000 to January 2016. We
of data like pottery, chipped stones, figurines, fauna, etc., searched in the databases by using the following keywords:
can use statistics as a tool to clarify its hypotheses and “Archeology”, “multivariate analysis”, “anthropology”,
scientific expectations. Advanced statistical methods, e.g. “complicated statistical tools” with “OR” and “AND” and
multivariate statistical methods, are commonly applied in “NOT” Boolean Operators in the Title/Abstract/Keywords
extensive inter-related collections. field (Table 1). Then the articles were controlled in order
Each statistical method should be applied to certain to be consistent with inclusion and exclusion criteria of
areas of archaeological studies; it is important to extract the research. Preferred Reporting Items for Systematic
biological, economic, social and cultural information Reviews and Meta-Analyses (PRISMA) as a powerful and
properly via statistical methods. For example, multivariate a common checklist in systematic reviews has been used
statistical methods and cluster analysis are commonly for assessing the quality of the articles selected for this
used in classification of archaeological sites which have research. (http://www.prisma-statement.org).
most resemblances based on the measured variables; on The inclusion criteria in the selection of related articles
the other hand, they could be applied in investigations are:
of cultural interactions. Principle Component Analysis is - Period of publication: from January 2000 to January
mostly applied in classification of cultural data (pottery, 2016.
coins, fauna, etc.) from archaeological sites. - Language: English.
The basic statistical tools described above provide for - Databases: PubMed, Scopus, Science Direct.
fundamental, quantitative description and comparison, - Methods: using a multivariate statistical method includ-
for establishing the confidence with which the samples ing cluster analysis, factor analysis, principal components
available permit characterizations of the populations analysis, discriminant analysis, multidimensional scaling
from which they come, and for assessing the strength and multivariate multiple regression.
and significance of the relationships between pairs of These articles were excluded based on the exclusion cri-
variables. Advanced statistical methods are also often teria:
used in archaeology: cluster analysis, factor analysis, - Irrelevant articles.
principal components analysis, discriminant analysis, - Duplicate articles.
multidimensional scaling, multivariate regression, and - Articles that lacked full-text.
others (Drennan 2009). - Conference papers.
Following Binford and Clarke, statistical analyzes After application of inclusion and exclusion criteria,
were applied extensively by European and North finally 384 articles were selected for this investigation. Then
American researchers, but unfortunately statistical the details of the statistical methods and their application in
methods were not developed in archaeological studies archaeological research was extracted (details are shown in
by Near Eastern archaeologists. In Iran, except for a few the Figure 1).
89
International Journal of the Society of Iranian Archaeologists Vol. 2, No. 4, Summer- Autumn 2016
Fig. 1: Bibliographical search and inclusion process (PRISMA diagram).
To assess the quality of the articles, PRISMA check from articles including:
list was used. The list has 27 items and each item has a - Authors specifications.
score or rating. PRISMA can assess all parts of an article - Publication year.
including the title, abstract, introduction, methods, results - Journal specifications.
as well as discussion and conclusion. Each part of an article - Multivariate statistical methods applied.
could be assessed by maximum points in PRISMA and is - Cultural material under investigation.
assessed separately, for example the title has a maximal of All this information was documented in a database with
1 point, the abstract 1 point, the introduction 2 points, the proper structure in Excel (Table 2 in the Appendix).
method 12 points, the results 7 points and for discussion Table 1 shows the search strategy of global databases
and conclusion there are 4 points. Hence and article could in this research. In other words, it clarifies how inclusion
be assessed from 0 points to a maximum of 27 points. In and exclusion criteria have been applied for filtering period
this research, article with less than 15 points have been and language and other criteria to select specific articles
excluded due to unacceptable quality. properly. Using this search strategy makes it possible to
Next, each author investigated the articles separately use advanced search box in each of global databases for
and independently and extracted the required information finding related articles.
90
Vol. 2, No. 4, Summer- Autumn 2016 International Journal of the Society of Iranian Archaeologists
Table 1: Search strategy
Search strategy in PubMed/ Scopus/Science Direct
(("Archeology"[MeSH Terms] OR “anthropology“ [MeSH Terms]) AND (“multivariate analysis” [ MeSH Terms ]
OR “complicated statistical tools” [Text Word]) OR ("cluster analysis"[MeSH Term]) OR "factor analysis"[ MeSH Term])
OR "principal components analysis"[MeSH Term]) OR "discriminant analysis"[MeSH Term]) OR "multidimensional
scaling"[Text Word]) OR "multiple regression"[MeSH Term]) OR "multivariate regression"[MeSH Term])) AND
language ( English ) AND ( Filters: Publication date from 2000/01/01 to so far )
Results
and Humanities” and are under publications of Springer
Using the strategy described above, a total of 4520 articles and Elsevier. More than 30% of them are from American
were retrieved from the three databases mentioned. 2120 of Scholars.
them were excluded because of irrelevant titles, abstracts All of the 384 articles were classified based on
and keywords. In the next step, 1220 articles were excluded multivariate statistical methods and then the application of
due to non-English language or the unavailability of their these methods in archaeology and cultural material types
full texts. From 1180 remaining articles, 292 articles were was determined. In addition, more than one multivariate
excluded because their publication year was earlier or later method was applied in some articles. Table 3 and Figure
than the specific period of this research (i.e. 2000 - 2016); 2 present the total number and relative frequency of
another 465 articles were excluded because they were application of each multivariate statistical method in
not in the index of three databases (i.e. Scopus, Science archaeology. They show that methods, including Cluster
Direct and PubMed); 34 more articles were excluded due analysis, Principal Component Analysis, Discriminant
to doubleness. Finally, after application of inclusion and Analysis, Multivariate Multiple Regression, Factor
exclusion criteria, 384 remaining articles were selected for Analysis and Multidimensional Scaling have had
the systematic review of this research (Figure 1). respectively the highest application in archaeological
Table 2 presents the information from some of the investigations. In addition, Cluster Analysis (as a method
articles, including authors’ specifications, titles, journals, of data reduction) was the most applied method in
publication year, etc. Articles concerning statistical clustering archaeological sites based on similarities in
methods in archaeology are mostly published in 2013 and cultural materials; also, Principal Component Analysis
the least of them are published in 2000. The results indicate (as a dimensionality reduction method) was mostly
that application of multivariate statistical methods in applied in dimensionality reduction of cultural materials
archaeology increased from 2009 and culminated in 2013. from archaeological sites. Various cultural materials have
More than 50% of the article reviewed in this research have been analyzed in these articles and include human bones,
been published after 2010. Most of these articles are from fauna, pot sherds, chipped stones, flora and archaeological
Journal of Archaeological Science in the field of “Arts structures.
Table 2: Descriptive statistics for any type of multivariate approach.
Multivariate approach The total number of each method have been used in the whole of 384 articles
Cluster Analysis 124 (32.4%)
Factor Analysis 21 (5.4%)
Principal Component Analysis 104 (27%)
Discriminant Analysis 83 (21.6%)
Multidimensional Scaling 10 (2.7%)
Multivariate Multiple Regression 42 (10.9%)
* All of cells reported according to frequency (percentage %)
91
International Journal of the Society of Iranian Archaeologists Vol. 2, No. 4, Summer- Autumn 2016
Fig. 2: Graphical display of percentages of multivariate approaches.
Discussion and Conclusion Most of multivariate statistical methods are applied
in surface surveys and there are a limited number of
There are two subsets of statistics in archaeology (Niknami excavations in which the researcher has used Multivariate
2012): Multiple Regression for prediction. In surface survey and
1) Primary statistics for describing various cultural sampling, multivariate approaches are used in comparisons
material, e.g. chipped stones, fauna, bones, etc. between collections (of chips, stones, organic materials,
2) Analytical statistics for deduction, comparison and ceramics, etc.) from different sites or comparison and
interpretation of archaeological data and their structures. interpretation of site structures (Drennan 2009; Craig et al.
Although primary statistics have always been more used 2006).
in description of cultural material (Froehle et al. 2012), in This study aims to encourage archaeologists to use
recent decades’ application of advanced statistical methods advanced statistical methods, esp. multivariate methods
in analytical research has notably increased; especially in archaeological research. Advanced statistical methods
advanced multivariate statistical methods have been could have a wide variety of applications. They have
particularly applied since 2009, indicating the significance various applications in archaeology as well, as in most
of statistical methods in archaeological research covering areas of archaeological research from surface survey to
description to interpretation. Advanced statistical methods, excavation and archaeological interpretation.
one of the most important of which is multivariate The results of this systematic review indicate that
methods, helps the researchers to improve their analysis cluster analysis is one of the most applied statistical
and interpretation of individual to collective and help with analysis, perhaps because of usage, method and simple
the descriptions by considering errors in measurements interpretation, compared to other methods of data
(Stutz and Estabrook 2004). reduction (Kovarovic et al. 2011). This method is used for
92
Vol. 2, No. 4, Summer- Autumn 2016 International Journal of the Society of Iranian Archaeologists
data reduction or clustering archaeological sites based on from the Long-Glassow Collection, University of New Mexico Press,
Albuquerque, pp. 67-100.
their similarities and helps with the comparison between
site structures (Papageorgiou and Liritzis 2007; Hart and Drennan, R. D.,
Matson 2009). 2009 Statistics for Archaeolgists: A Common Sense Approach. 2nd
ed, Springer, New York.
Principle Component Analysis is the second most
widely used methods due to its application in any data El Shazly, E. M.,
structure and simplicity of interpretation compared to 1957 Classification of Egyptian Mineral Deposits. Egyptian Journal
of Geology 1 (1), 1-20.
other methods of dimensionality reduction. Methods of
dimensionality reduction are mostly applied in the cultural Froehle A.W.; C. M. Kellner and M. J. Schoeninger,
material classification and categorizing archaeological 2012 Multivariate Carbon and Nitrogen Stable Isotope Model for
the Reconstruction of Prehistoric Human Diet. American Journal of
material from surface sampling and simple interpretation Physical Anthropology 147 (3), 352-369.
of archaeological sites. It seems that in future research
more data reduction methods and dimensionality reduction Gowland R. L. and A. G. Western,
2012 Morbidity in the Marshes: Using Spatial Epidemiology to
methods would be applied, because more surface surveys Investigate Skeletal Evidence for Malaria in Anglo-Saxon England (AD
than excavations are encouraged. 410-1050). American Journal of Physical Anthropology 147 (2), 301-311.
Hart J. P. and R. G. Matson,
Acknowledgement: 2009 The Use of Multiple Discriminant Analysis in Classifying
Prehistoric Phytolith Assemblages Recovered from Cooking Residues.
The authors are greatly thankful to the director and principal Journal of Archaeological Science 36 (1),74-83.
of the Tarbiat Modares University for their constant Kovarovic K., L. C. Aiello, A. Cardini and C. A. Lockwood,
encouragement and support for this study. We are also 2011 Discriminant Function Analyses in Archaeology: Are
thankful to all the faculty members of “Archeology and Classification Rates Too Good to Be True? Journal of Archaeological
Science 38 (11), 3006-3018.
Biostatistics” for the technical assistance in the research.
The authors would also like to thank Dr. Mozhgan Jayez Myers, O. H.,
for her kind support during this investigation. 1950 Some Applications of Statistics to Archaeology. Government
Press, Cairo.
Conflict of interest: Niknami. K.,
2012 Advanced Statistical Methods in the Analysis of Archaeological
Data. Samt, Tehran (In Persian).
All authors declare that there is no conflict of interests in
the current study. Papageorgiou I. and I. Liritzis,
2007 Mumtivariate Mixture of Normals with Unknown Number of
Components: An Application to Cluster Neolithic Ceramics from Aegean
Funding: and Asia Minor Using Portable XRF. Archaeometry 49 (4), 795-813.
No Funding. Rezalo, R.,
2009 Application of Statistics in Archaeological Researchs. MA
Thesis, Mohaghegh Ardabily University (in Persian).
Stutz A.J and G. F. Estabrook,
References 2004 Computationally Intensive Multivariate Statistics and Relative
Frequency Distributions in Archaeology (With An Application to the
Baxter M. J., Early Epipaleolithic of the Levant). Journal of Archaeological Science
2009 Archaeological data analysis and fuzzy clustering. 31(12), 1643-1658.
Archaeometry 51 (6), 1035-1054.
Walker P. L.,
Craig N., M. Aldenderfer and H. Moyes, 2008 Sexing Skulls Using Discriminant Function Analysis of
2006 Multivariate Visualization and Analysis of Photomapped Visually Assessed Traits. American Journal of Physical Anthropology
Artifact Scatters. Journal of Archaeological Science 33(11), 1617-1627. 136 (1), 39-50.
de Leeuw, J., Xiancen, L. and X. Renping,
2007 Correspondence Analysis of Archeological Abundance 1995 Application to Mathematical Statistics to Data Processing of
Matrices. In: Nance, C.R.; De Leeuw J.; Prado, K. and D. Verity (Eds.), Scientific and Technological Archeology. Journal of Guangxi University
Correspondence Analysis and West Mexico Archaeology: Ceramics for Nationalities (Natural Science Edition) 1 (2), 63-68.
93
View publication stats
