Validating Theories Through Analytic Generalization
1. 1
ANALYTIC GENERALIZATION:
VALIDATING THEORIES THROUGH RESEARCH BY
MANAGEMENT PRACTITIONERS AND STUDENTS
Michael Pasco, DVM, DBA, San Beda University, Philippines
Abstract
For the management practitioners and students, the number of cases in management research was not a
hindrance to contribute valuable studies. Because of the limited published specific guidance, I reviewed
related literature about analytic generalization and its importance to theory testing in synergy to the
experiences and access of the researcher to target units of analysis and cases. I introduced pattern matching,
triangulation method and case descriptions as important techniques to perform analytic generalization. I
also reviewed that analytic generalization was complimentary to data and conclusions gained through
statistical generalization. While both qualitative research and quantitative research had strengths and
weaknesses, the use of mixed methods were available options that maximize the evaluation and making
conclusions with both analytic generalization and statistical generalization. I reviewed that the ways to
present results through tables, graphs and figures to test hypothesis and arrive at analytic generalization
were important to be designed. The theoretical framework of strengths and weaknesses of qualitative and
quantitative research methodologies (Choy, 2014) summarized the rationale for the choice of research
methods by the researcher.
Keywords: Analytic generalization, Pattern matching, Triangulation method
Introduction
I noticed that there are many studies about valuable management practices, processes, structure
or techniques that are not pursued by the researchers, managers and students because of the perceived
limitations of one or few number of cases, such as firms or organizations, when thinking of statistical
generalization in making research conclusions. I reviewed related literature to validate research gaps
and suggest rigorous alternatives to quantitative research and statistical generalization. I also
highlight techniques to enhance writing of qualitative research and mixed research methods.
Polit and Beck (2010) defined âgeneralization is an act of reasoning that involves drawing
broad conclusions from particular instancesâthat is, making an inference about the unobserved
based on the observedâ (p. 1451-1452).
Ostlund, Kidd, Wengstrom, and Dewar (2011) view that âthere is a lack of pragmatic guidance
in the research literature as how to combine qualitative and quantitative approaches and how to
integrate qualitative and quantitative findingsâ (p. 370). I review the characteristics of analytic
generalization, statistical generalization, triangulation method, and mixed methods. In situations
where there is limited number of cases or respondents, what are the considerations for management
practitioners and students to design and complete a relevant research?
Analytic Generalization
Perry (2000) describes analytic generalization as a method that involves formerly developed
theory used as a template and describes the empirical results about the relationships between
variables. The template is then used to compare with the observed results from a case (Perry, 2000).
Yin (2009) suggests that analytic generalization fits case study research method. If a case or
more cases validate the same theory, replication can be claimed (Yin, 2009). The researchers are
concerned with replication logic of the cases and theory rather than the sampling logic (Yin, 2009).
2. 2
Yin (2009) asserts that case studies are evaluated based on construct validity (concepts being studied
are operationalized and measured correctly), internal validity (rigorous causal relationships), external
validity (generalizability of findings outside immediate case), and reliability (demonstration of
repeatability of results from same methods).
The values of construct validity, internal validity, external validity and reliability of research
are also viewed by Eisenhardt (1989), Trochim (1989) and Yin (2009). The practical use and rigor of
descriptive theory building fit the investigation of different valuable phenomena in firms or
organizations (Trochim, 1989; Nicholson & Kiel, 2007; Yin, 2009).
In analytic generalization, Firestone (1993) notes that researchers strive to generalize from particulars
to broader constructs or theory. Analytic generalization is not just linked with qualitative research but is
embedded within theory-driven quantitative research (Firestone, 1993).
In an idealized model of analytic generalization, qualitative researchers develop
conceptualizations of processes and human experiences through in-depth scrutiny and higher-order
abstraction (Firestone, 1993). In the course of their analysis, qualitative researchers distinguish
between information that is relevant to many study participants, in contrast to aspects of the
experience that are unique to particular participants (Firestone, 1993). Generalizing to a theory or
conceptualization is a matter of identifying evidence that supports that conceptualization with
richness in descriptions (Firestone, 1993).
Similar with statistical generalizability, analytic generalization is an ideal model that is not
always realized (Polit & Beck, 2010). Meredith (1998; p. 448) described case study and analytic
generalization as applicable in ânatural controlled observation, logic as controlled deduction, theory
for replicability and theoretics for generalizability.â
Pattern matching. Yin (2009) also views pattern matching as one of the specific strategies to
come up with analytic generalization. Pattern matching compares an empirically based pattern with a
predicted one (Yin, 2009). If the patterns coincide, the results can strengthen the internal validity of
the case study (Yin, 2009).
The types of pattern matching are nonequivalent dependent variables as a pattern, rival
explanations as patterns and simpler patterns (Yin, 2009). In nonequivalent dependent variables,
quasi-experiment may have multiple dependent variables or variety of outcomes (Perry, 2000). If, for
each outcome, the initially predicted values have been found, and at the same time alternative
âpatternsâ of predicted values have not been found, strong causal inferences can be made (Yin, 2009,
Perry, 2000). With rival explanations, each case has certain type of outcome, and the investigation
has to be focused on how and why this outcome occurred (Yin, 2009; Perry, 2000).
This analysis demands the development of rival theoretical propositions, stated in operational terms
(Perry, 2000). The rival explanation has a pattern of independent variables that is mutually exclusive
(Yin, 2009; Perry, 2000). If one explanation is to be valid, the others cannot be (Yin, 2009; Perry,
2000).
If there are two different dependent (or independent) variables, simpler pattern is possible as
long as a different pattern has been stipulated for these two variables (Yin, 2009; Perry, 2000). The
lesser the variables, the more dramatic the different patterns will have to allow any comparisons of
their differences (Yin, 2009; Perry, 2000).
Analytic generalization through pattern matching is used to compare research propositions with
observed data (Trochim, 1989; Nicholson & Kiel, 2007; Yin, 2009). Data analyzed and theories were
validated with triangulation method (Yeung, 1997; Yin 2009). As an option, a researcher can analyze
with the pattern matching of modal quantitative and qualitative responses, non-metric categorical
3. 3
data for descriptions of contexts, ordinal data from survey, interview of research participants, and
participant observations for analyses of associations between variables.
Outcome pattern matching. Trochim (1989) defined outcome pattern matching for
generalization âthat required a theoretical pattern of expected outcomes, an observed pattern of
effects, and the attempt to match the twoâ (p.360)
Table 1
Interpretation of pattern matching of ordinal data
Independent Variable Dependent Variable Interpretation
High High Complete match
Low Low Complete match
High Low Partial match
Low High Partial match
High Absent No match
Low Absent No match
Absent High No match
Absent Low No match
Absent Absent No match
I propose an example of the implementation of pattern matching that can be used for research,
as seen in Table 1. To initiate, the expected outcomes related to variables and theories have to be
defined. For instance, positive relationship (High level of Independent Variable and High level of
Dependent Variable; or Low level of Independent Variable and Low level of Dependent Variable)
meant âcomplete matchâ or literal replication of the propositions in the research conceptual
framework (Trochim, 1989; Nicholson & Kiel, 2007; Yin, 2009). Inverse relationship also means
âpartial matchâ or partial replication of an appropriate proposition in this research as this shows
influence to dependent variables (Trochim, 1989; Nicholson & Kiel, 2007; Yin, 2009, Jaccard &
Jacoby, 2010). However, zero association meant âno matchâ or no replication of appropriate
proposition (Trochim, 1989; Nicholson & Kiel, 2007; Yin, 2009; Jaccard & Jacoby, 2010).
Interpretation of the findings required iteration between propositions and observed data and
matched sufficiently contrasting rival patterns to the data (Yin, 2009; Ritchie & Lewis, 2003;
Nicholson & Kiel, 2007; Eisenhardt, 1989).
Case descriptions. Some of the data that can be analyzed through pattern marching are
categories, contexts of responses, frequency, percentage, mode of responses, and distinguished
perspectives from respondents (Eisendhardt, 1989). The case descriptions can be written similar to
features stories or informative reports related to the profile, background, industry, best practices and
activities of the case (Yin, 2009). Most of the information must be relevant to the research topic (Yin,
2009).
Case study research. Yin (2009) recommends analytic generalization for case study research
and suggests that case studies are appropriate where the objective is to study contemporary events,
and where it is not necessary to control behavioral events or variables. Yin (2009) further suggests
single case study research is also appropriate if the objective of the research is to explore new
subject, description of phenomenon, theory building, or theory testing.
4. 4
Triangulation Method
Jick (1979; p. 602) cited Denzin (1978) who broadly defined triangulation âas the combination
of methodologies in the study of the same phenomenon.â Multiple reference points to locate an exact
position and multiple viewpoints for greater accuracy (Jick, 1979).
Torrance (2012) observed that:
Over the last ten years or so, the âFieldâ of âMixed Methods Researchâ (MMR) has
increasingly been exerting itself as something separate, novel and significant, with some
advocates claiming paradigmatic status. Triangulation is an important component of
mixed method designs. Triangulation has its origins in attempts to validate research
findings by generating and comparing different sorts of data, and different respondentsâ
perspectives, on the topic under investigation. (p.1)
Types of triangulation. Yin (2009) cited Patton (2002) that the 4 types of triangulation are
data (multiple data sources), investigator (among different evaluators), theory (using different
perspectives), and methodological (using different methods).
Convergence of data. Yin (2009) also identified the importance of convergence of multiple
sources of evidence to come up with research conclusion like âdocuments, archival records, open-
ended interviews, focus interviews, structured interviews and surveys, and observations (direct and
participant)â (p. 117).
Triangulation is defined as the use of multiple methods mainly qualitative and quantitative
methods in studying the same phenomenon for the purpose of increasing study credibility. This
implies that triangulation is the combination of two or more methodological approaches,
theoretical perspectives, data sources, investigators and analysis methods to study the same
phenomenon. (Hussein, 2009; p. 2)
Researchers also confirm the association between variables of interests through documents and
participant observations that note down the observable data and evidence (Eisenhardt, 1989;
Trochim, 1989; Nicholson & Kiel, 2007; Yin, 2009). Coghlan & Brannick (2014) recommend each
researcherâs reflection as a source of data for analysis in action research.
Mixed Research Methods
Ostlund et al. (2011; p. 369) notice that there is a trend for conducting parallel data analysis on
quantitative and qualitative data in mixed methods research. Using triangulation as a methodological
metaphor, the researchers clarify their theoretical propositions and the bases of their results (Ostlund
et al., 2011). Mixed methods âoffer a better understanding of the links between theory and empirical
findings, challenge theoretical assumptions and develop new theoryâ (Ostlund et al, 2011; p. 369).
Ostlund et al. (2011) assert further that mixed methods research, where quantitative and
qualitative methods are combined, is increasingly recognized as valuable, because it can potentially
capitalize on the respective strengths of quantitative and qualitative approachesâ (p. 369).
In contrast, Hussein (2009) considered:
However, using both qualitative and quantitative paradigms in the same study has
resulted into debate from some researchers arguing that the two paradigms differ
epistemologically and ontologically. Nevertheless, both paradigms are designed towards
understanding about a particular subject area of interest and both of them have strengths
and weaknesses. Thus, when combined there is a great possibility of neutralizing the
flaws of one method and strengthening the benefits of the other for the better research
results. (p.2)
Hussein (2009) also views that triangulation is a good way to gain the benefits of both
5. 5
qualitative and quantitative methods. The use of triangulation however will depend on the
researcherâs philosophical choice of the kinds and types of research. Malina, Norreklit, and Selto
(2010) suggest that the divides between quantitative and qualitative methods as well as economic and
more behavioral theories are not constructive toward understanding management phenomena.
Creswell (1999) suggest that mixed methods enable the policy makers to understand complex
phenomena while presenting the number and graphs that are quite understandable. Triangulation
improves the generalizability of the research results (Creswell, 1999).
Borrego, Douglas, & Amelink (2009) identified the triggers for the selection of either
quantitative, qualitative or mixed research methods. These factors are the research problem, the
personal experiences of the researcher, and the access to audience or respondents.
Presentation of Findings for Analytic Generalization
Tabular approach. This approach has to be designed and learned. Results to be used in
analytic generalization are presented using tables to describe or explain each proposition or
hypothesis (Yin, 2009). These tables are preferably to have columns and provided each for data
results of pattern matching, narrative descriptions of results, participant observations, documents,
and/or other multiple sources of evidence (Yin, 2009; Nicholson & Kiel, 2007). Eisenhardt (1989)
implied the effective use of tabular approach to present and consolidate quantitative and qualitative
data results gathered in mixed research methods. Statistical data were supported by qualitative data
descriptions that are meaningful and relevant to the research outcomes (Eisenhardt, 1989).
Nicholson and Kiel (2007) effectively utilized tabular approach to evaluate pattern matching
that compared expected and observed ordinal data. The information corroborated to present research
findings (Yin, 2009; Nicholson & Kiel, 2007).
Graphs and Figures. Nicholson and Kiel (2007) also presented results in figures. Trochim
(1989) introduced outcome pattern matching while he effectively portrayed research outputs using
graphs and figures. Indeed, there were clear presentations of the similarities or difference of the
patterns of the relationships between variables in either graphs or figures (Trochim, 1989; Anderson
et al., 2015).
Strengths and Weaknesses of Quantitative Research
I view that quantitative research is the norm. Choy (2014) viewed that quantitative research can
be administered and evaluated quickly. Numerical data are more comparative and has strong reliability
(Choy, 2014).
In contrast, Choy (2014; p. 102) cited Dudwick, Kuehnast, Jones, & Woolcock (2006) that
âmany important characteristics of people and communities including both rich and poor, for example,
identities, perceptions, and beliefs are not meaningfully reduced to numbers or adequately understood
without reference to the local context in which people live.â
I observed that many management researchers and research students have difficulty in accessing
to numerous firms and organizations to gather empirical observations. Large sample sizes are also
expected from quantitative research to come up with statistical generalization (Choy, 2014).
Statistical Generalization
Anderson, Sweeney, & Thomas (2015) defines statistical inference as a process of data from
samples to estimate or test hypothesis about a population. Perry (2000) defines statistical
generalization as making an inference about a population on the basis of empirical data collected
about a sample and it is a level one inference.
6. 6
Statistical inferences are also âbases of methods we use when we generalize from observed
dataâ (Freund & Perles, 2007; p. 3). This method of generalization is commonly recognized because
research investigators have quantitative formulas characterizing generalizations that can be made
using significance level, confidence level, size of the effect, power of test (Perry, 2000). But, using
statistical generalization as a method of generalizing the results of a case study is a flaw because
cases are not sampling units, nor shall they be chosen for this reason (Perry, 2000).
The best strategy for achieving a representative sample is to use probability (random)
methods of sampling, which give every member of the population an equal chance to be
included in the study with a determinable probability of selection. Standard tests of
statistical inference are based on the assumption that random sampling from the target
population has occurred. Like most models, this generalizability model is an idealâa goal
to be achieved, rather than an accurate depiction of what transpires in real-world research.
Yet the myth that this model is adhered to in quantitative scientific inquiry in the human
sciences perseveres. (Polit & Beck, 2010; p. 1452-1453)
Polit and Beck (2010) believe that the ârandom sampling myth is one in which virtually all
researchers conspire when they apply standard statistical tests to analyze their data, in violation of the
assumption of random samplingâ (Polit & Beck, 2010; p. 1453).
Strengths and Weaknesses of Qualitative Research
Comparatively, the weakness of qualitative research is associated with qualitative cultural
analysis as time-consuming, and the researchers interpretations are limited (Yauch and Steudel, 2003;
as cited by Choy, 2014). As positioned subjects, the personal experience and knowledge influence the
observations, possible bias, and conclusions (Yauch and Steudel, 2003; as cited by Choy, 2014).
Figure 1. The model of strengths and weaknesses of qualitative and quantitative research
methodologies (Choy, 2014)
7. 7
Because qualitative inquiry can be considered as open-ended, the research participants have
more control over the content of the data collected (Yauch and Steudel, 2003; as cited by Choy, 2014).
With different scope, qualitative methods support researchers to explore the views of uniform as
well as diverse groups of people help unpack these differing perspectives within groups (Choy, 2014).
The primary strength of the qualitative approach to cultural assessment is the ability to probe for
underlying values, beliefs, and assumptions (Choy, 2014), as seen in Figure 1.
Summary
For the management practitioners and students, the number of cases in management research is
not an impediment to contribute valuable and relevant research. There are limited published specific
guidance in articles to combine qualitative and quantitative approaches and how to integrate qualitative
and quantitative findings and conclusions. Through review of related literature, I described analytic
generalization as important to validate theory with synergy to the experiences and access of the
researcher to target units of analysis and cases. Pattern matching, triangulation method and case
descriptions are important techniques to perform analytic generalization. Analytic generalization can
be complimentary to data and conclusions gained through statistical generalization. While both
qualitative research and quantitative research have strengths and weaknesses, the use of mixed
methods are available options that maximize the use of both analytic generalization and statistical
generalization. The ways to present results through tables, graphs and figures to test hypothesis and
arrive at analytic generalization are important to be learned and designed. The theoretical model of
strengths and weaknesses of qualitative and quantitative research methodologies (Choy, 2014)
summarizes the rationale for the choice of research methods and techniques by the researcher.
References
Abowitz, D. A., & Toole, T. M. (2010). Mixed method research: Fundamental issues, of design,
validity, and reliability in construction research. Journal of Construction Engineering and
Management, 2010. Doi: 10.1061/(ASCE).1943-7862.0000026
Anderson D. R., Sweeney, D. J., & Thomas, W. A. (2015). Modern Business Statistics (4th ed.).
Pasig City, Philippines: Cengage Learning Asia Pte Ltd.
Borrego, M., Douglas, E. P., & Amelink, C. T. (2009). Qualitative, quantitative, and mixed research
methods in engineering education. Journal of Engineering Education, (2009), 53-66. Retrieved
from https://pdfs.semanticscholar.org/ab21/6e664b2dff79bd44044f6b9e78e58bfa2f02.pdf
Choy, L. T. (2014). Strengths and weaknesses of research methodology: comparison and
complimentary between qualitative and quantitative approaches. IOSR Journal of Humanities
and Social Science (IOSR-JHSS), 19(4), 99-104. Retrieved from http://www.iosrjournals.org
Coghlan, D., & Brannick, T. (2014). Doing action research in your own organization (4th
ed.).
London: SAGE.
Creswell, J. W. (1999). Mixed method research: Introduction and application, In Chapter 18.
Handbook of Educational Policy. USA: Academic Press. Retrieved from
http://cachescan.bcub.ro/e-book/V/580599_6.pdf
Eisenhardt, K. M. (1989). Building theories from case study research. The Academy of Management
Review, 14(4), 532-550. Retrieved from: http://www.jstor.org/
8. 8
Firestone, W. A. (1993). Alternative arguments for generalizing from data as applied to qualitative
research. Educational Researcher, 22(4), 16-23. Retrieved from
http://www.jstor.org/stable/1177100
Freund, J. E., & Perles, B. M. (2007). Modern Elementary Statistics (12th
ed.). Jurong, Singapore:
Pearson.
Hussein, A. (2009). The use of triangulation in social research: can qualitative and quantitative
methods be combined? Journal of Comparative Social Work, 1, 1-12. Retrieved from:
http://journal.uia.no/index.php/JCSW/article/viewFile/212/147
Jick, T. D. (1979). Mixing qualitative and quantitative methods: Triangulation in action.
Administrative Science Quarterly, 24, 602-611. Retrieved from http://business.illinois.edu/
Malina, M. A., Norreklit, H. S. O., & Selto, F. H. (2011). Lessons learned: Advantages and
disadvantages of mixed method research. Qualitative Research in Accounting and
Management, 2011. Retrieved from https://www.researchgate.net/publication/227430195
Meredith, J. (1998). Building operations management theory through case and field research. Journal
of Operations Management, 16, 441-454. Retrieved from https://pdfs.semanticscholar.org/
Nicholson, G. J., & Kiel, G. C. (2007). Can Directors impact performance? A case based test of three
theories of corporate governance. Corporate Governance: An International Review, 15(4), 585-
608. Retrieved from http://eprints.qut.edu.au
Ostlund, U., Kidd, L., Wengstrom, Y., & Dewar, N. R. (2011). Combing qualitative and quantitative
research within mixed method research designs: A methodological review. International
Journal of Nursing Studies, 48, 369-383. Doi: 10.1016/j.ijnurstu.2010.10.005
Perry, C. (1998). Processes of a case study methodology for post-graduate research in marketing.
European Journal of Marketing, 32(9-10), 785-802. Retrieved from https://www.bedicon.org/
Perry, D. E. (2000). Case Studies. Unpublished Presentation. Retrieved from
http://users.ece.utexas.edu/~perry/education/382c/L06.pdf
Polit, D. F., & Beck, C. T. (2010). Generalization in quantitative and qualitative research: Myths and
strategies. International Journal of Nursing Studies, 47, 1451-1458. Doi:
10.1016/j.ijnurstu.2010.06.004
Torrance, H. (2012). Triangulation, respondent validation, and democratic participation in mixed
methods research. Journal of Mixed Methods Research, 6(2), 111-123. Retrieved from http://e-
space.mmu.ac.uk/302804
Trochim, W. M. K. (1989). Outcome pattern matching and program theory. Evaluation and Planning
Program, 12, 355-366. Retrieved from
https://www.socialresearchmethods.net/research/Outcome%20Pattern20Matching%20and%20
Program%20Theory.pdf
Yin, R. K. (2009). Case study research (4th
ed.). London: SAGE.