Download Causal Comparative and Co-relational Research and more Study notes Advanced Education in PDF only on Docsity!
Causal Comparative and Co-relational Research Objectives, Rational, Critiques Designs: Causal Comparative and Correlation Designs Challenges Assumptions that guide our development of knowledge Assumptions about reality (O___?) What is the nature of reality? Something “real” and discoverable? Constructed and co-created? Channeled and imposed? Assumptions about knowledge (E___?) How do individuals and societies pursue and substantiate knowledge? Assumptions about values (A___?) What are the valued ends of our search for knowledge? How should we deal with different valued ends of knowledge in the conduct of research? (Ontology ( ontos- from the Greek, meaning being, is) this domain asks “what is?” “what exists?” “what is the nature of reality?” Epistemology ( episteme - from the Greek, meaning knowledge) this domain examines what our theory of knowledge is and asks “how do we know we know?” “what is the nature of knowledge?” Axiology ( axios- from the Greek, meaning worth) this domain examines what the highest value or good is and asks “What is the good?” “What are the valued ends of research?” “Where do the researcher’s values belong?”) Research Paradigms… and Research Assumptions Paradigms: Rough groupings of ontological, epistemological and axiological assumptions that shape our thinking and approaches to inquiry … (Neo) positivism Interpretivism/Phenomenology Constructionism/Narrative Inquiry Emancipatory/Transformative Postmodernism/Pragmatic Positivism: The Elephant as an Object of Study O__: Reality exists “out there” apart from my awareness (realism) E__: Through study and experiment, I will discover how reality works “out there” (objectivist) A__: Through technical understanding I can shape and control reality to serve my interests
(For the positivist/postpositivist paradigm, the elephant is an object. The realist ontology tells us that the elephant-object has inherent meaning, that meaning lies within the elephant. The objectivist epistemology suggests that we can come to “discover” the meaning in the elephant-object, what the elephant/object is through rigorous and precise measurement, observation, dissection, disassembly, reassembly. This will provide us with universal (generalizable) laws that we can apply to elephants everywhere as the valued end of positivist/postpositivist research is prediction and control so that we can manage the elephant, grow the elephant, master the elephant. The researcher assumes a value-free stance because the interest is in establishing facts, there is no place for the researcher’s values in the research.) Experimental & Quasi-Exp Research Objectives: Pinpoint a well-specified cause-effect relationship in a sample that will “generalize” to a known population Rationale: If we seek to shape social processes and outcomes in particular ways, experimental research that is both internally and externally valid contribute to efficient and effective control Threats Relationship not valid (Internal validity) Relationship does not generalize (External validity) (“Recent federal education policies (e.g., the No Child Left Behind [NCLB] Act of 2001 [NCLB, 2001] and the Education Sciences Reform Act [ESRA] of 2002 [ESRA, 2002]) have generated considerable debate among education researchers. Much of the debate has been about what is meant by “scientifically based research” in these policies and the implications for research in education. Some researchers are pleased to see experiments, meta-analyses, and randomized trials cited as exemplary methods of scientifically based research; from their perspective, these powerful research methods are not currently being used to good advantage in education research (e.g., Slavin, entifically based” research in education? and Is scientifically based research
pages of Educa- tional Researcher (Jacob & White, 2002).” (p. 31)) Experimental & Quasi-Exp Research Critiques and Concerns Feasibility: Is it possible to design and conduct? What are the costs of establishing the needed controls? Applicability: Do design choices to enhance internal validity reflect a poor trade-off in terms of external validity? (Sidebar on Occam’s Razor) How substantially have experimental conditions deviated from those in the population at large? Ethics: Is it appropriate to subject human participants (as individuals or groups) to experimental conditions? Have they consented? Is the consent informed? Is the treatment or outcomes humane? Fair? Holy Threats to Validity, Batman! Considerations of Feasibility, Applicability and Ethical Responsibility come into play as the experimenter grapples with threats to: Internal validity (Did I pinpoint the relationship in the sample?) External validity (Will the relationship found in the sample generalize to the population?) Efforts to gain control over the experiment and the participants come at costs These costs may erode feasibility, applicability and or ethical responsibilities Single Case Designs …“Closely follows the logic laid out…for experimental and quasi-experimental research…” (Mertens, p. 207) Direct interventions, direct observations, long duration, repeated and multiple measures In this sample, N=1 so we cannot generalize findings. Note: “applied to several people” and “treatment should be standardized” (p.
- imply replication Note: Social validity (p. 212) suggests attention to axiology that extends beyond technical control Despite deviations from “the gold standard,” Single Cases still shine light While single case studies adhere in many ways to the logics of experimental and quasi-experimental designs… By design, the sample is NOT random Given the design, causal relationships CANNOT be isolated and determined Given the design, findings CANNOT be generalized to the population
… But the study of such cases can be feasible, may be applicable to a broader population, and ought to be ethical in treatment of the subject Causal Comparative & Correlational Research Objectives: What are the primary objectives of Causal Comparative and Correlational research? Research in these forms seek to establish the presence and strength of relationships among characteristics found in society Rationale: In many circumstances, CC & C research present viable options for research. What are these? Experimental designs may be infeasible, impractical and/or unethical. CC& C designs can provide lower cost means to study social processes and outcomes as they occur rather than under conditions of experimental control Causal Comparative Research Comparison Group (e.g. parent education level) Pre-existing groups/not randomly assigned Groups: hs dropout, hs grad, college grad) Compared on some known characteristic (college adm) No intervention/No manipulation of characteristic Note status: Admitted to college or not) Presence of a statistically significant relationship Suggests a possible cause. Presence in comparison group ( presumed cause) associated with characteristic of interest ( presumed effect). Determination of a causal relationship would demand experimental research (Confidence comes at a cost!) Causal Comparative Research: Value Relative to experimentation, a low-cost means to establish a relationship exists (even if causation not established) May identify presence of a relationship and encourage experimentation (small initial investment to narrow the field) Well-established procedures (Chi-Square, ANOVA, MANOVA, etc.) and knowledge base help to lower the costs of design Correlational Research: Types Prediction Studies: Seeks to determine if the presence/strength of a prior characteristic (e.g. early test performance) help to predict a later characteristic (e.g. graduation or income). Relationship Studies: Seeks to determine the presence/strength of a relations b/w co-existing characteristics
Recall: Correlational Research Can identify presence of a relationship and encourage experimentation Relative to experimentation, a low-cost means to establish a relationship Developed techniques are robust and powerful Well-established procedures (Multiple Regression, Logistic Regression, Factor Analysis, Discriminant Analysis) and new procedures (HLM, SEM) Holy Batpoop, Lucius! Regression sounds cool. Say more… Social scientists have developed techniques to mirror some of the “control” characteristic of experimental designs in correlational designs Multiple regression analysis can reveal correlations between a dependent variable and a variety of independent variables These models give “statistical” control of confounding variables to reveal relationships b/w variables of interest These models are more or less robust, but depend on data/measures meeting a number of conditions There are a variety of models developed for particular purposes/settings Multiple Regression
y= ax + by + cz + …+ error term y =dependent variable; x, y, z =ind variables; R^2 =(explained variance) =power of the “model” a,b,c, …= “correlation coefficients” which indicate the relationship b/w that independent variable and the dependent variable, “controlling” for the effects of the other variables. (Thinks of these as tuning “knobs”). Cor coefficients may or may not be statistically significant Simplified example Survey of US workforce Salary=a(yrs of exp)+b(college)+c(gender)+error R 2 = 52%; a =0.03, b=0.3, c= - 0. CC&C Research: Critiques and challenges While CC&C presents a viable option for research, these are subject to limitations (see M pp. 153-9) B/C independent variables are not strictly controlled, the presence of a relationship does not necessarily imply it is causal. (Correlations is not causation!) In relying on pre-existing characteristics, CC & C may struggle with “unpacking” loaded group labels (e.g. sex v. gender, mixed race categories, etc.) Methodological and publication bias overemphasize difference and underplay commonality As tools emerge to process complex relationships (HLM, SEM, etc.), both the power to discern and risks associated with misperceiving relationships grow
