analytic features of LCA that had been in serious development since the
1950s. Despite the expansive presentation and motivation for LCA
provided by Lazarsfeld and Henry (1968), there were still two primary
barriers to larger scale adoption of latent clas
extended in various ways according to the substantive interests and
empirical demands of differenr disciplines as well as the varying
curiosities of quantitative methodologists, statisticians, biosratisticians,
psychometricians, and econometricians. As su
positivism holds that the methods and procedures of the natural
sciences are appropriate to the social sciences and also maintains (the
ontological position) that there is one objective reality independent of
the observer. Methodologically, positivism is
OXFORD LIBRARY OF PSYCHOLOGY Editor-in-Chief PETER E. NATHAN The
Oxford Handbook of Quantitative Methods Edited by Todd D. Little
VoLUME 2: STATISTICAL ANALYSIS OXFORD UNIVERSITY PRESS OXFORD
UN lVI!RSlTY PRESS l )xfiml University Press is a department of
stakeholders, and rapid feedback mechanisms that lead to corrective
action or dissemination of best practice. PM&E is a learning process,
where stakeholders learn, on the one hand, to develop and adjust
methods and techniques for evaluation, negotiation,
2) the environment in which the project is to take place is promising but
uncharted; 3) to pilot a potentially good development effort when the
time is right and the ground has already been laid through a local
initiative, social assessment, other partici
discrete response variables with classspecific multinomial distributions.
However, LCA has a rich history within tl1e psychometric tradition,
somewhat independent of the development of finite mixture models,
that is worthy of remark, not unlike the way in
Piloting Phase Follow-up Phase 28 7. Conclusion This note has outlined
the factors critical for evaluating the impacts of projects and programs.
Three conclusions emerge: 1. Identification of the counterfactual is the
organizing principle of a good impact
binary items, strong associations with a particular class or high class
homogeneity is indicated by high or low model-estimated probabilities
of endorsement-char is, Wntlft or I-wmlkclose I, with "close" defined by
Wmlk > .7 or Wmik < .3. For example, con
instruments have a strong monitoring and evaluation component. In the
case of LILs, monitoring and evaluation (M&E) are an explicit part of the
loan, and rapid feedback from M&E is essential to success. In the case
of APLs, M&E is essential for the borrow
America by Oxfiml University Press 198 M<-Statistical methods. 2.
Psychology-Mathematical models. I. Litde, Todd D. BF39.0927 2012
150.721 1-dc23 2012015005 9 8 7 6 5 4 3 2 Printed in the United States
of America on acid-free paper CHAPTER 25 Latent Class
Decentralized. Washington, D.C.: World Bank. (Mimeo). Rossi, Peter H.
and H. E. Freeman. 1982. Evaluation: A Systematic Approach. Second
Edition. California: SAGE Publications, Inc. 30 Appendix Note 1
Qualitative and quantitative methods roots and differe
theory approach, develops theoretical ideas from observations of the
data themselves, by constructing sensitizing concepts from
observation, drawing out comparisons with other linked areas and
sample theoretically, e.g., by sampling critical cases. This a
latent profile analysis (LPA), in terms of model assumptions,
specification, estimation, evaluation, selection, and interpretation. In
addition, a brief introduction to structural equation mixture modeling in
the form of latent class regression is provide
and program justification based on broadly acknowledged economic
and other parameters for the program; 3) a viable financing plan and a
general description of the required investments and activities which
would be defined in due course; 4) as the program
distribution functions of the response variable, conditional on latent
class membership, f(heightic = 1), . ,f(heightjc = K). Recognizing
mixture models as latent variable models allows use of the discourse
language of the latent variable modeling world.
patterns is tantamount to grouping individuals, this framing ofLCA is
more person-oriented. Thus, in both the psychometric tradition in
which LCA was developed and in the classical mathematical statistics
tradition in which finite mixture modeling was dev
concepts both relate to the parameters of the unconditional
measurement model and ultimately qualifY the interpretation of the
resultant latent classes, consider a hypothetical example with five 558
LATENT CLASS ANALYSIS AND FINITE MIXTURE MODELING binary
quantitative/statistical) that will facilitate the estimation of a proxy
variable to capture these effects. 3. Both quantitative and qualitative
methods are necessary for a good evaluation. The two approaches
strongly complement each other. This note has
undersl'allding of the region. That is, the region itself doesn't change
but the information that can be gleaned about the region does change
according to the type of search, and determining which search is more
useful depends entirely on the objccrives o
item response probabilities for each class along with the item response
odds ratios calculated following Equation 7. Classes 1, 2, and 3 all have
high homogeneity with respect to items UJ, u2, and u4 because all
class-specific item response probabilities
analysis, are personcentered. Variable-cenrered approaches describe
associations rlmong variables and are predicated on the assumption
that the population is homogeneous with respect to the relationships
between variables (Laursen & Hoff, 2006, p. 379). I
participatory techniques can be quantified. The Bank distinguishes
different levels of participation: information dissemination (one-way
flow of information); consultation (two-way flow of information);
collaboration (shared control over decision making);
Pftmale are referred to as the mixingproportions and finale (height) and
/rernate(height) are the component distribution density functions. The
second common feature for all the different kinds of mixture models
previously listed is that the components th
see McLachlan and Peel (2000) and Vermunr and McCutcheon (2012).
For more recent developments specifically related to LCA, see
Hagenaars and McCutcheon cfw_2002) and Collins and Lanza (20 1 0).
There has also been conspicuous growth in the number of stati
traditional factor analysis, conditional or local independence is assumed
for the M items conditional on class membership. This assumption
implies that latent class membership explains all of the associations
among the observed items. Thus, the formation
indirect applications is then nor on the resultant mixture components
nor their interpretation bur, rather, on the overall population
distribution approximated by the mixing. I find the indirect versus direct
application distinction for mixture modeling l
distributions, fi (height), . ,fK (height), are all unknown but can be
estimated, under certain identifYing assumptions, using height data
measured on a representative sample from the total population. Finite
Mixture Models As Latent Variable Models It is
corresponding to the actual observed response patterns. Essentially, in
the observed data there are a maximum of 1024 groupings of
individuals based on their observed responses. Latent class analysis
then enables the researcher to group or cluster these r
of one or more variables as a mixture of or composite of a finite
number of component distributions, usually simpler and more tractable
in form than the overall distribution. As an example, consider the
distribution of adult heights li1 the general popula