See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/278671692

Multivariate Analysis
Chapter January 2007
DOI: 10.13140/RG.2.1.4326.9283

CITATIONS

READS

50

2 authors, including:
Nina Baur
Technische Universitt Berlin
198 PUBLICATIONS 311 CITATIONS
SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The Social Construction of and Interaction in Space by Using Maps View project

Process-Oriented Methodology View project

All content following this page was uploaded by Nina Baur on 18 June 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.

3120

multivariate analysis

scope is pandisciplinary. Scholars have used the

concept of race, class, and gender as an interlocking site of oppression, in multiple ways, to
create theory as an analytical tool or as a methodological practice (Berger 2004). Research explicitly utilizing intersectional analysis tends to
cluster in pockets in a few traditional social
science disciplines (sociology, psychology, education) and in multidisciplinary programs including
womens studies, ethnic studies, criminology,
and environmental studies. Several sociologists
have compiled anthologies that examine the
intersections of race, class, and gender. Two key
texts that provide a conceptual framework for
understanding the complex intersections of
oppressions have been written by sociologists:
Privilege, Power, and Difference (2001) by Allan
Johnson and Understanding Race, Class, Gender,
and Sexuality: A Conceptual Framework (2001) by
Lynn Weber.
Multiracial feminism is a burgeoning field
that centers on the voices of women of color
but includes writings by anti-racist white
women, women outside the US, and feminist
men of color. Comprehending the intersections
of oppressions in order to promote equity
across lines of race, class, and gender and
nation differences is a key component of multiracial feminism.
Sociologists have contributed greatly to this
endeavor. Multiracial feminism offers new formulations about organizing, coalition building,
and critical theory production. The field has
reached a maturity and sophistication in both
activist and scholarly communities, enriching
the conceptualization of power, identity, and
inequality.

Anzaldua, G. (Ed.) (1990) Making Face, Making

Soul/Haciendo Caras: Creative and Critical Perspectives by Feminists of Color, 1st edn. Aunt Lute
Foundation Books, San Francisco.
Bambara, T. C. (Ed.) (1970) The Black Woman: An
Anthology. Washington Square Press, New York.
Berger, M. (2004) Workable Sisterhood: The Political
Journey of Stigmatized Women with HIV/AIDS.
Princeton University Press, Princeton.
Brant, B. (Ed.) (1984) A Gathering of Spirit: A Collection by North American Indian Women. Firebrand Books, Ithaca, NY.
Hurtado, A. (1996) The Color of Privilege: Three
Blasphemies on Race and Feminism. University of
Michigan Press, Ann Arbor.
Mohanty, C. T., Russo, A., & Torres, L. (Eds.)
(1991) Third World Women and the Politics of Feminism. Indiana University Press, Bloomington.
Moraga, C. & Anzaldua, G. (Eds.) (1983) This Bridge
Called My Back: Writings by Radical Women of
Color, 2nd edn. Kitchen Table Women of Color
Press, New York.
Sandoval, C. (1990) Feminism and Racism: A Report
on the 1981 National Womens Studies Association Conference. In: Anzaldua, G. (Ed.), Making
Face, Making Soul/Haciendo Caras: Creative and
Critical Perspectives by Feminists of Color, 1st edn.
Aunt Lute Foundation Books, San Francisco,
pp. 5571.
Sandoval, C. (2000) Methodology of the Oppressed.
University of Minnesota Press, Minneapolis.
Smith, B. (Ed.) (1983) Home Girls: A Black Feminist
Anthology. Kitchen Table Women of Color Press,
New York.
Thompson, B. (2001) A Promise and a Way of Life:
White Antiracist Activism. University of Minnesota
Press, Minneapolis.
Wong, D. & Cachapero, E. (Eds.) (1989) Making
Waves: An Anthology of Writings By and About
Asian American Women. Beacon Press, Boston.

SEE ALSO: Black Feminist Thought; Feminism; Feminism, First, Second, and Third
Waves; Intersectionality; Matrix of Domination; Race; Race (Racism); Third World and
Postcolonial Feminisms/Subaltern; Transnational and Global Feminisms

multivariate analysis

REFERENCES AND SUGGESTED

READINGS
Anzaldua, G. (1987) Borderlands/La Frontera: The
New Mestiza. Aunt Lute Foundation Books, San
Francisco.

Nina Baur and Siegfried Lamnek

One way to classify quantitative methods of
social research is by the number of variables
involved in statistical data analysis procedures.
In univariate statistics, just one variable is of
interest, while in bivariate statistics the relation
between two variables is analyzed. Multivariate
analysis involves, in the loose sense of the term,

multivariate analysis
more than two variables and, in its strict sense,
at least two dependent and two independent
variables. With increasing numbers of variables,
statistical modeling becomes necessary and
more complex. At the same time, these models
are more appropriate for social sciences, since
in social reality many variables are intertwined
and there is rarely one central determination.
Once data are collected and read into a database processable by statistical software, the
typical steps in a multivariate data analysis are
the following.
1 Framing the research question in such a way
that it can be modeled mathematically.
2 Selecting the right statistical model, since
many kinds of multivariate methods exist
and researchers continually develop new
multivariate methods. Every multivariate
model searches for certain patterns in data.
It might miss other patterns. Using different multivariate methods therefore may
lead to different results. Thus selecting
the right theoretical model at the beginning
of data analysis is essential. For example, a
statistical correlation may point to a
causal relation or a latent variable. If
one applies regression analysis, one usually
only investigates the possibility of a causal
relation. However, experienced researchers
can use the same statistical method for different theoretical goals, e.g., regression analysis could also be used in other ways than
causal analysis.
3 Verifying that assumptions and prerequisites
for the chosen statistical procedure are
met. Most multivariate procedures require
at least (a) valid, standardized data; (b) a
minimum number of cases; (c) a random
sample; (d) a specified variable scale type;
(e) a certain (very often, a normal) distribution of variables and residuals; (f ) a minimum variance of variables; (g) in causal
analysis usually independence of independent variables. If any of these assumptions
are not met, a different multivariate procedure should be chosen as results may be
erroneous. Again, profound research
experience and statistical knowledge are
needed to assess when violated assumptions
lead to invalid results and when they lead
only to less stable results.

3121

4 Preparing data for the specific analysis. A

special case of data preparation is data
mining, which specializes in extracting variables from complex data banks for statistical
analysis.
5 Computing the model using a special statistical computer package such as SAS, SPSS,
or Stata.
6 The results of data analysis always have to be
interpreted. Statistics may help in interpreting data, but they never prove theories.
Multivariate analysis procedures can be classified in different ways, and no classification is
exhaustive, especially due to the dynamics of
the field. Two common ways are (1) on what
scale variables have to be measured in order to
apply the model and (2) what kind of theoretical model underlies the analytic procedure.
Among the theoretical questions multivariate
analysis can address are (a) identifying latent
classes; (b) causal analysis; (c) identifying patterns in time; (d) network analysis; and (e)
multilevel analysis. Most multivariate procedures can be viewed as a special case of general
linear models (GLM).
Identifying latent classes. Associated variables
can be interpreted measuring a background
variable that was not or cannot be measured,
such as typologies, classes, or dimensions. For
example, the correct number of answers in a
test can be seen as a sign of greater or lesser
intelligence. Esping-Andersen identified welfare regimes by classifying countries according
to socioeconomic and political similarities.
Multivariate analysis procedures that identify
latent classes include cluster analysis, correspondence analysis, factor analysis, principal
component analysis, and multidimensional scaling (MDS).
Causal analysis. Correlation can also mean
that one or more variables induce one or more
other variables. For example, education and
country of origin both have a strong influence
on income. Note the specific underlying concept of causality: cause in the sense of statistical techniques usually means the
relationship between variables, while, as Abbott
(2001) points out, only persons (not the variables used to describe them) can act and thus
cause anything. Most statistical techniques

3122

multivariate analysis

Table 1

Types of cases and groups.

Independent variables

Dependent variables

Nominal

Metric

Nominal

Metric

Correspondence analysis
Log-linear models
Tree analysis, e.g. CHAID
Analysis of variance
(ANOVA; MANOVA)

Discriminant analysis
Logistic regression

of causal analysis try to assess existence, kind,

and strength of causal relationships between
variables. One distinguishes, for example, nonrecursive relationships (i.e., one-way causal relationships); recursive relationships (variable A
influences variable B and vice versa); additive
multicausality (many causes independently
affect the dependent variable); interaction (a
cause only impacts if the case belongs to a certain category in a third control variable);
common causes (one dependent variable influences several distinct independent variables);
intervention (a causal chain exists, i.e., variable
A influences variable B, which in turn influences variable C, etc.); circularity (variable A
influences variable B; variable B causes variable
C; variable C affects variable A). Causal links
between variables can be much more complex,
especially if variable number increases. Most
social scientists focus on establishing the kind
of causal relationship between many variables
and on distinguishing causally relevant variables
from irrelevant variables. Table 1 gives an
(incomplete) overview of multivariate procedures for establishing causal relations, classified
by the minimum variable scale required. If few
relevant variables (preferably on a metric scale)
exist, one can try to estimate the exact effect size
using econometric techniques.
Identifying patterns in time. Many methods
exist to analyze the change of (typical) human
action. Most of them originally stem from
demography, economic sociology, and life
course research. Examples are cohort analysis,
times series, event-history analysis (survival
analysis), latent growth curve models, and

Canonical regression
Partial correlation
Multiple regression
Multivariate regression
Path analysis (LISREL)

sequential analysis (optimal matching techniques). These methods usually either require a
variable measuring time (such as age) or
research designs with several measuring points
(such as trend design, panel design, continuous
measuring).
Network analysis. An individual case (such as
a person, a word, situation) can interact with
other individuals but is often part of higherlevel cases forming a collective ( aggregate),
e.g., a person can be part of an organization or a
country; a word can be an element of a sentence
or a book; a situation can be part of an interaction system. Network analysis procedures investigate the relation between individuals forming a
collective on a higher level, for example the
intensity of social contacts between members
of a non-governmental organization.
Multilevel analysis. Sometimes, the relationship between different analysis levels is of interest, e.g., the influence of regional unemployment
rate (analysis level: region) on voting behavior
(analysis level: persons) or the effect of youth
violence (analysis level: persons) on legislation
(analysis level: countries). In this case, multilevel
analysis procedures are applicable.
SEE ALSO: ANOVA (Analysis of Variance);
Computer-Aided/Mediated Analysis; Experimental Methods; Factor Analysis; General
Linear Model; Hypotheses; Latent Growth
Curve Models; Log-Linear Models; Mathematical Sociology; Path Analysis; Quantitative
Methods; Regression and Regression Analysis;
Social Network Analysis; Statistics; Structural
Equation Modeling; Time Series; Variables

Mumford, Lewis (18951990)

REFERENCES AND SUGGESTED
READINGS
Abbott, A. (2001) Time Matters: On Theory and
Method. Chicago University Press, Chicago.
Abbott, A. & Tsay, A. (2000) Sequence Analysis and
Optimal Matching Methods in Sociology: Review
and Prospect. Sociological Methods and Research
29: 333.
Blasius, J. & Greenacre, M. (2004) Correspondence
Analysis in the Social Sciences: Recent Developments
and Applications. Academic Press, London.
Blossfeld, H.-P. & Rohwer, G. (2002) Techniques of
Event History Modeling: New Approaches to Causal
Analysis. Erlbaum, Mahwah, NJ.
Han, J. & Kamber, M. (2006) Data Mining: Concepts
and Techniques. Morgan Kaufmann, San Francisco.
Kennedy, P. (2003) A Guide to Econometrics. MIT
Press, Cambridge, MA.
Scott, J. & Xie, Y. (Eds.) (2005) Quantitative Social
Science, 4 vols. Sage, London.
Snijders, T. A. B. & Bosker, R. J. (1999) Multilevel
Analysis: An Introduction to Basic and Advanced
Multilevel Modeling. Sage, London.
StatSoft Inc. (19842005) Electronic Textbook
StatSoft. Online. www.statsoft.com/textbook/
stathome.html.

Mumford, Lewis
(18951990)
Mark Luccarelli
Lewis Mumford was born in New York City.
He is best known as an architectural critic and
urban historian, and author of The City in
History (1961), undoubtedly his greatest work.
Of mixed German and German-Jewish heritage, Mumford grew up with his mother and
maternal grandfather an immigrant and head
waiter who took the boy for long walks, initiating Mumfords lifelong interest in cities. With
the publication of The Culture of Cities (1938)
Mumford achieved widespread recognition that
grew when the New Yorker magazine hired him
to write its Skyline column. A journalist by
profession, he had wide-ranging intellectual
interests and wrote convincingly on a variety
of topics, including technology and culture,
literary criticism, social ethics, and politics.

3123

Mumford was a public intellectual who

addressed the educated public rather than a
specific academic community. Despite or perhaps because of his audience, he undertook an
ambitious intellectual project: the reexamination of modernity in the light of growing
scientific and imaginative understandings of
evolutionary processes. Mumford thought that
in respect to developments in technology,
architecture, and urban form, culture could be
likened to the workings of nature, and like the
philosopher John Dewey he held that creative
innovation holds the key to evolving designs
appropriate to the task of reconciling human
values and natural processes.
The barbarism of World War I and the subsequent collapse of Progressive Era reform politics in the US set the stage for Mumfords
intellectual journey. A polymath, he naturally
took to Emersons maxim that all education is
essentially autobiographical. Aside from diverse
courses he took at the City College of New
York that failed to add up to a Bachelors
degree, his most important training came from
contacts he developed himself, particularly during a formative trip to Britain undertaken in the
1920s. These contacts included the botanist and
urbanist Patrick Geddes and the sociologist
Victor Branford. From Geddes, Mumford
learned the technique of the regional survey
of the city and its environs. Through Branford,
he was introduced to sociological theory and in
particular the ideas of the nineteenth-century
French regionalist, Frederick Le Play. Regionalism became a philosophy of living and an
expression of Mumfords concern for the
dynamics of place. As visiting editor of Branfords journal Sociological Review, Mumford
published a series of articles on regionalism in
which he argued that an empirical understanding of both the natural and man-made environment is fundamental to the assessment of
culture. Natural geographic patterns contribute
to fundamental ecologies that shape culture,
regardless of the level of sophistication (technological and economic) achieved. Mumford
was primarily interested in regional ecologies;
that is, in the technological and aesthetic principles that underlie a built environment and
structure its relation to the surrounding natural
region. At the same time, he reasoned along the
lines of the Chicago School of sociology, that