- The chapter discusses experimental and quasi-experimental research designs. The classical experiment tests the effect of an experimental stimulus through pretesting and posttesting experimental and control groups. Random assignment helps make groups comparable. Experiments allow tight control but results may lack generalizability. Quasi-experiments are used when true experiments are not possible. Time series designs and case studies examine variables in few cases. Designs can be customized for different research purposes.
Research Design, Causation, Measurement in Criminal Justice
1. Chapter 4 : General Issues in Research Design
Learning Objectives:
1. Recognize how explanatory scientific research centers on the notion of cause and effect,
and why this is a probabilistic model of causation.
2. Describe the three basic requirements for establishing a causal relationship in science,
together with what is a necessary cause and a sufficient cause.
3. Understand the role of validity and threats to validity of causal interference.
4. Summarize the four classes of validity threats, and how they correspond to questions about
cause and effect.
5. Discuss how a scientific realist approach bridges idiographic and nomothetic approaches to
causation.
6. Describe different units of analysis in criminal justice research.
7. Explain how the ecological fallacy relates to units of analysis.
8. Understand the time dimension, together with the differences between cross-sectional and
longitudinal research.
9. Describe how retrospective studies may approximate longitudinal studies.
Summary:
• Explanatory scientific research centers on the notion of cause and effect.
• Most explanatory social research uses a probabilistic model of causation. X may be said to
cause Y if Y always happens when X happens.
• Three basic requirements determine a causal relationship in scientific research: (1) The
independent variable must occur before the dependent variable, (2) the independent and
dependent variables must be empirically related to each other, and (3) the observed relationship
cannot be explained away as the effect of another variable.
• When scientists consider whether causal statements are true or false, they are concerned with
the validity of causal interference.
• Four classes of threats to validity correspond to the types of questions researchers ask in trying
to establish cause and effect. Threats to statistical conclusion validity and internal validity arise
from bias. External and construct validity threats may limit our ability to generalize from an
observed relationship.
• A scientific realist approach to examining mechanisms in context bridges idiographic and
nomothetic approaches to causation.
• Units of analysis are the people or things whose characteristics researchers observe, describe,
and explain. The unit of analysis in criminal justice research is often the individual person, but
it may also be a group, organization, or social artifact.
• Researchers sometimes confuse units of analysis, resulting in the ecological fallacy or the
individual fallacy.
• Cross-sectional studies are those based on observations made at the same time. Although such
studies are limited by this characteristic, inferences can often be made about processes that
occur over time.
• Longitudinal studies are those in which observations are made at many times. Such
observations may be made of samples drawn from general populations (trend studies), samples
2. drawn form more specific subpopulations (cohort studies), or the same sample of people each
time (panel studies).
• Retrospective studies can sometimes approximate longitudinal studies, but retrospective
approaches must be used with care.
Key Terms:
Cohort study (Page 100)
Construct validity (Page 89)
Cross-sectional study (Page 99)
Ecological fallacy (Page 95)
External validity (Page 89)
Internal validity (Page 88)
Longitudinal study (Page 100)
Panel study (Page 100)
Probabilistic (Page 85)
Prospective (Page 102)
Retrospective (Page 101)
Scientific realism (Page 92)
Statistical conclusion validity (Page 87)
Trend study (Page 100)
Units of analysis (Page 92)
Validity (Page 87)
Validity threats (Page 87)
Chapter 5: Concepts, Operationalization, and Measurement
Learning Objectives:
1. Understand the role of concepts as summary devices for bringing together observations and
experiences that have something in common.
2. Explain how concepts are mental images that do not exist in the real world.
3. Describe how operationalization specifies concrete empirical procedures for measuring
variables.
4. Recognize the operationalization beings with study design but continues through the
duration of research.
5. Explain why measurement categories must be mutually exclusive and exhaustive.
6. Distinguish different levels of measurement and the properties of different levels.
7. Understand precision, reliability, and validity as dimensions of measurement quality.
8. Summarize how creating specific, reliable measures may not reflect the complexity of the
concepts we seek to study.
9. Understand how multiple measures of a concept can improve reliability and validity.
10. Describe composite measures and explain the advantages.
Summary:
3. • Concepts are mental images we use as summary devices for bringing together observations and
experiences that seem to have something in common.
• Our concepts do not exist in the real world, so they can’t be measured directly.
• In operationalization, we specify concrete empirical procedures that will result in
measurements of variables.
• Operationalization brings in study design and continues throughout the research project,
including the analysis of data.
• Categories in a measure must be mutually exclusive and exhaustive.
• Higher levels of measurements specify categories that have ranked order or more complex
numerical properties.
• A given variable can sometimes be measured at different levels of measurement. The most
appropriate level of measurement used depends on the purpose of the measurement.
• Precision refers to the exactness of the measure used in an observation or description of an
attribute.
• Reliability and validity are criteria for measurement quality. Valid measures are truly indicators
of underlying concepts. A reliable measure is consistent.
• The creation of specific, reliable measures often seems to diminish the richness of meaning our
general concepts have. A good solution is to use multiple measures, each of which taps
different aspects of the concept.
• Composite measures, formed by combining two or more variables, are often more valid
measures of complex criminal justice concepts.
Key Terms:
Concepts (Page 111)
Conception (Page 110)
Conceptual definition (Page 115)
Conceptualization (Page 112)
Construct validity (Page 128)
Content validity (Page 128)
Criterion-related validity (Page 127)
Dimension (Page 112)
Face validity (Page 127)
Interval measures (Page 121)
Nominal measures (Page 121)
Operational definition (Page 115)
Ordinal measures (Page 121)
Ratio measures (Page 122)
Reification (Page 113)
Reliability (Page 124)
Typology (Page 132)
Validity (Page 127)
Chapter 6: Measuring Crime
Learning Objectives:
4. 1. Recognize how different approaches to measuring crime illustrate general principles of
conceptualization, operationalization, and measurement.
2. Understand what crimes are included in different measures.
3. Describe measures of crime and how they are based on different units of analysis.
4. Understand different purposes for collecting crime data.
5. Explain different measures based on crimes known to police.
6. Describe the main features of victim surveys.
7. Distinguish the main differences between crimes known to police and crimes measured
through different types of surveys.
8. Understand why self-report measures are used, and list different types of crimes for which
they are appropriate.
9. Summarize major series of self-reported measures of drug use.
10. Understand how surveillance measures of crime safety are obtained and used.
11. Explain how different measures of crime satisfy criteria for measurement quality.
12. Recognize that we have different measures of crime because each measure is imperfect.
Summary:
• Crime is a fundamental concept in criminal justice research. Different approaches to measuring
crime illustrate the general principles of conceptualization, operationalization, and
measurement.
• Before using any measure of crime, researchers should understand what types of offenses it
does and does not include.
• Different measures of crime are based on different units of analysis. The UCR is a summary
measure that reports totals for for individual agencies. Other measures use offenders, victims,
incidents, or offenses as the units of analysis.
• Crime data are collected for one or more general purposes: monitoring, agency accountability,
and research.
• Crimes known to police have been the most widely used measures. UCR data have been
available since the early twentieth century; more detailed information about homicides was
added to the UCR in 1961. Most recently, the FBI has developed an incident-based reporting
system that is gradually being adopted.
• Surveys of victims reveal information about crimes that are not reported to police. The NCVS
includes very detailed information about personal and household incidents but does not count
crimes against businesses or individual victims under age 12. Although the NCVS is a
nationally representative measure, it cannot estimate victimization for states or local areas.
• Self-report surveys were developed to measure crimes with unclear victims that are less often
detected by police. Two such surveys estimate drug use among high school seniors and adults.
Self-report surveys do not measure all drug use because of incomplete reporting by respondents
and procedures for selecting survey respondents.
• ADAM II and DAWN provide measures of drug use among special populations but are best
suited to monitoring changes in drug use.
• Different measures of crime are also developed for specific research and policy purposes.
Many police departments do crime analysis with their own incident-based records.
5. • We have many different measures of crime because each measure is imperfect. Each measure
has its own strengths and weaknesses.
Key Terms:
Crimes known to police (Page 142)
Dark figure of unreported crime (Page 149)
Incident-based measure (Page 146)
Self-report survey (Page 155)
Summary-based measure (Page 146)
Surveillance system (Page 142)
Victim Survey (Page 149)
Chapter 7: Experimental and Quasi-Experimental Designs
Learning Objectives:
1. Recognize that experiments are well suited for the controlled testing of causal processes
and for some evaluation studies.
2. Describe how the classical experiment tests the effect of an experimental stimulus on some
dependent variable through the pretesting and posttesting of experimental and control
groups.
3. Understand that a group of experimental subjects need not be representative of some larger
population but that the experimental and control groups must be similar to each other.
4. Describe how random assignment is the best way to achieve comparability in the
experimental and control groups.
5. Describe how the classical experiment with random assignment of subjects guards against
most of the threats to internal validity.
6. Understand that the controlled conditions under which experiments take place may restrict
our ability to generalize results to real-world constructs or to other settings.
7. Recognize how the classical experiment may be modified by changing the number of
experimental and control groups, the number and types of experimental stimuli, and the
number of pretest or posttest measurements.
8. Know the reasons that quasi-experimental are conducted when it is not possible or
desirable to use an experimental design, and be able to describe different categories of
quasi-experiments.
9. Understand the differences between case-oriented and value-oriented research. Time-series
designs and case studies are examples of variable-oriented research, in which a large
number of variables are examined for one or a few cases.
10. Be able to describe how experiments and quasi-experiments can be customized by using
design building blocks to suit particular research purposes.
Summary:
• Experiments are an excellent vehicle for the controlled testing of causal processes. Experiments
may also be appropriate for evaluation studies.
6. • The classical experiment tests the effect of an experimental stimulus on some dependent
variable through the pretesting and posttesting of experimental and control groups.
• It is less important that a group of experimental subjects be representative of some larger
population than that experimental and control groups be similar to each other.
• Random assignment is the best way to achieve comparability in the experimental and control
groups.
• The classical experiment with random assignment of subjects guards against most of the threats
to internal validity.
• Because experiments often take place under controlled conditions, results may not be
generalizable to real-world situations, or findings from an experiment in one setting may not
apply to other settings.
• The classical experiment may be modified to suit specific research purposes by changing the
number of experimental and control groups, the number and types of experimental stimuli, and
the number of pretest or posttest measurements.
• Quasi-experiments may be conducted when it is not possible or desirable to use an
experimental design.
• Nonequivalent-groups and time-series designs are two general types of quasi-experiments.
• Time-series designs and case studies are examples of variable-oriented research, in which a
large number of variables are examined for one or a few cases.
• Both experiments and quasi-experiments may be customized by using design building blocks to
suit particular research purposes.
• Not all research purposes and questions are amenable to experimental or quasi-experimental
designs because researchers may not be able to exercise the required degree of control.
Key Terms:
Case-oriented research (Page 192)
Case study (Page 192)
Classical experiment (Page 169)
Control group (Page 170)
Dependent variable (Page 169)
Experimental group (Page 171)
Generalizability (Page 179)
Independent variable (Page 169)
Quasi-experiment (Page 183)
Random assignment (Page 173)
Variable-oriented research (Page 192)
7. • The classical experiment tests the effect of an experimental stimulus on some dependent
variable through the pretesting and posttesting of experimental and control groups.
• It is less important that a group of experimental subjects be representative of some larger
population than that experimental and control groups be similar to each other.
• Random assignment is the best way to achieve comparability in the experimental and control
groups.
• The classical experiment with random assignment of subjects guards against most of the threats
to internal validity.
• Because experiments often take place under controlled conditions, results may not be
generalizable to real-world situations, or findings from an experiment in one setting may not
apply to other settings.
• The classical experiment may be modified to suit specific research purposes by changing the
number of experimental and control groups, the number and types of experimental stimuli, and
the number of pretest or posttest measurements.
• Quasi-experiments may be conducted when it is not possible or desirable to use an
experimental design.
• Nonequivalent-groups and time-series designs are two general types of quasi-experiments.
• Time-series designs and case studies are examples of variable-oriented research, in which a
large number of variables are examined for one or a few cases.
• Both experiments and quasi-experiments may be customized by using design building blocks to
suit particular research purposes.
• Not all research purposes and questions are amenable to experimental or quasi-experimental
designs because researchers may not be able to exercise the required degree of control.
Key Terms:
Case-oriented research (Page 192)
Case study (Page 192)
Classical experiment (Page 169)
Control group (Page 170)
Dependent variable (Page 169)
Experimental group (Page 171)
Generalizability (Page 179)
Independent variable (Page 169)
Quasi-experiment (Page 183)
Random assignment (Page 173)
Variable-oriented research (Page 192)