Conceptualization, Operationalization, and Measurement

1
Chapter 5

• Conceptualization, Operationalization, and
  Measurement

2
Chapter Outline

• Measuring Anything That Exists
• Conceptualization
• Definitions in Descriptive and Explanatory
  Studies
• Operationalization Choices
• Criteria of Measurement Quality

3
Questions

• What does probably mean?
• Give the definition in a percentage
• How about old and young?
• Give a specific age range
• Define prejudice
• Use one sentence

4
Measuring Social Constructs

• How can we measure love, grief, prejudice,
  intelligence, and other non-quantifiable
  concepts?
• Social variables dont have any fixed meaning
• We define it specifically for the purposes of
  research
• You might each disagree with a static
  definition of intelligence, but a specific,
  measurable definition causes fewer problems
• This is the process of conceptualization

5
Conceptualization

• Process of specifying what we mean when we use
  particular terms
• Conceptualization is the process of creating
  concepts
• Concepts are constructs
• Constructs are constructions of meaning we
  create from our experience, but that have no
  concrete reality, no intrinsic meaning
• We individually give constructs meaning
• Theyre not real, but they have value (efficient)
• How would you conceptualize compassion?

6
Conceptualization

• Definitions are more a problem for descriptive
  research than for explanatory
• Reification is when we give concrete meaning to
  concepts (which are inherently abstract)
• Makes them useful, but causes problems
• Produces an agreed upon meaning for a concept for
  the purposes of research
• Includes descriptions of the indicators used to
  measure the concept and distinction between the
  different aspects of the concept (dimensions)

7
Conceptualization

• Indicators
• A sign of the presence or absence of a concept
• Allow us to quantify a concept like religious
• Dimensions
• There are many types or aspects of most
  concepts
• Example the organization, interpersonal,
  economic, philosophical dimensions of religious
  behavior

8
Definitions

• Real dictionary definition problematic in
  research because it mistakes a construct for a
  real entity (reification)
• The specification of concepts in the
  conceptualization process depends on
• Nominal definitions arbitrarily assigned to a
  term, without a claim that the definition
  represents a "real" entity. Usually has some
  level of agreement
• Operational definitions - Specifies how a concept
  will be measured in very clear, unambiguous
  language. This is the process of
  operationalization

9
From Concept to Measurement

• Progression from sense of what a term means to
  measurement in a scientific study
• Conceptualization
• Nominal Definition
• Operational Definition
• Measurements in the Real World
• This is a continual process, often a cycle
• Agreement about concept definition is not the
  goal the goal is create an absolutely specific
  and unambiguous definition
• Almost as though you are creating the term for
  the purposes of the study (sometimes you do)

10
Operationalization

• The development of specific research procedures
  that will result in empirical observations
  representing the concepts under study
• Range is an important consideration
• Age Number of months? Nearest year? Age range?
• You should try to accurately represent the range
  that your study needs
• Example is religious or not religious enough
  for your research purposes? Would more choices be
  better?
• When in doubt, get more than you need

11
Creation of Variables and Attributes

• Variables must have two qualities
• Attributes must be comprehensive (there should be
  no possibility that a subject cant answer the
  question)
• Education level must be more than finished
  high school only and finished college
• Attributes must be mutually exclusive (there
  should be no possibility that a subject will want
  to select more than one)

12
Four Levels of Measurement

• Nominal - offer names or labels for
  characteristics (gender, birthplace, telephone
  number, religion, etc.)
• You cant apply many statistical treatments to
  nominal data (e.g. mean is meaningless
• Nominal measures are only determined to be same
  or different
• Ordinal - variables with attributes we can
  logically rank and order
• One measure is identified as greater or less than
  another, but by no consistent amount
• Example level of uncertainty or satisfaction
  level

13
Four Levels of Measurement

• Interval consistent distances separate variable
  measures (temperature, IQ, deviation from mean
  height, etc.). Negative numbers possible
• No true zero point
• Ratio like interval, except attributes
  composing a variable are based on a true zero
  point (age, time, height, etc.). You can tell
  the ratio of one to the other (youre Dads twice
  as old as you)
• No negatives
• For many statistical analyses, there is no
  important difference between interval and ratio.
  It is often referred to as the interval/ratio
  level of measurement

14
Four Levels of Measurement

• Each ascending level tells us more
• Nominal different from another or same
• Ordinal different from another or same, more or
  less than another
• Interval different from another or same, more or
  less than another, how much more or less
• Ratio different from another, more or less than
  another, how much more or less, what is the ratio
  of one to the other
• Higher levels can be treated as lower ones (but
  lower ones cannot be converted to higher ones)

15
Measurement Quality

• Greater precision is usually preferable
• Precision is the fineness of an indicators
  distinctions
• Dont confuse precision and accuracy
• There are two primary considerations in research
  measurement
• Reliability
• Validity

16
Reliability

• Repeated measures should yield similar results
• A reliable measure offers no inherent assurance
  that youre measuring what you want to measure,
  just that they are consistent
• A single observer can cause reliability issues
• Tools to increase reliability
• Test-retest method do it more than once
• Split-half method use two equivalent measures on
  the same sample
• Use established measures someone else has done
  the work
• Use multiple coders, check the coders, train the
  coders

17
Validity

• Are we measuring what we think were measuring?
  (matches operational definition?)
• Measures should have face-validity
• Makes sense common agreement that the measure
  should be included (although not necessarily a
  complete measure)
• E.g. send your measure to experts for an opinion
• Measures should have criterion-related validity
• The extent that the measure correlates with
  other, external indicators or measures
• Example SAT, FCAT

18
Validity

• Measures should have construct validity.
• Your expectations of the construct are confirmed
• Does the measurement of the variable work the way
  you expect it to work?
• Does it seem as if you are measuring what you
  think you are measuring?
• Measures should have content validity it covers
  much of the range of possible meanings
• Does the measure include all relevant dimensions
  of the construct?
• See page 148, Figure 5-2

19
A Tradeoff

• Reliability and validity are a trade off
• Increased reliability often reduces validity
• Example highly reliable lab experiment is not
  likely to be as valid as studying individuals in
  their natural settings
• Increased validity often reduces reliability
• Studying individuals in a natural setting is
  valid (for those individuals), but replication is
  likely to offer different results

20
Example

• Is there a correlation between the closeness of a
  relationship and the communication of those in
  the relationship?
• Conceptualization?
• Nominal Definition?
• Operational Definition?