Concepts, Constructs, Variables, and Measurement

2. Concepts, Constructs, Variables, and Measurement

Concepts and Constructs

• Concepts are mental representations and are typically based on experience
• concepts can be of real phenomena (dogs, clouds, pain)
• concepts can be of agree-upon phenomena (truth, beauty, justice, prejudice, value, etc.)
• Three classes of things can be measured
• direct observables (height, weight, color, etc.)
• indirect observables (questionnaires provide information on gender, age, income, etc.)
• constructs (theoretical creations that are based on observations but which cannot be seen either directly or indirectly; things such as IQ, Leisure Satisfaction, Environmental Values, etc., are constructs

Variables and Measurement

• Measurement is the assignment of symbols to observable phenomena.
• There are two kinds of phenomena
• Constants
• Variables
• Concepts or constructs must be free to vary if they are to be variables; otherwise they are constants
• There are 3 ways of measuring things:
• counting
• ordering
• classifying
• Before variables can be measured they must be defined. Types of definitions:
• Theoretical: the words used in a theory; basically dictionary or common use
• Operational: a definition that explains how the variable is to be measured
• Operational definition: assigns a meaning to a concept or variable by specifying the operations needed to measure it
• Types of operational definitions
• directly measured: IQ, weight, attitude
• experimental: the details of how subjects are treated differently, such as: Aggressive behavior= banging toys, other children; frustration = what happens when children are in a room with toys they cannot reach.
• Independent and Dependent variables (except purely descriptive research)
• All research (except descriptive studies) must have at least two variables
• one can be IV and the other DV
• in symmetrical relationships, the question of which is independent and which is dependent is moot
• Having an IV allows you to assume a cause-effect relationship: changes in the IV result in changes in the DV
• If you cannot posit a cause-effect relationship, then you essentially have two IV's (the level/score of each is independent of the other [although both may depend on some other variable(s)])
• Having an IV allows more control and better inference about what is going on, especially when you have an active IV.
• Active and Attribute Independent Variables
• attribute: level or score of the variable is brought to the experiment by the subject, usually as a natural characteristic such as sex, age, etc.
• active: the level of the IV is manipulated by the experimenter
• Intervening variables: uncontrolled or unobserved variables that may account for variation in the DV (also known as extraneous variables)
• Control variables: any variable that may affect DV should be controlled; that is, measured and accounted for statistically or held constant (age, gender, socioeconomic status, etc., could be control variables)

Measurement levels of variables

• Nominal (qualitative)
• naming/classifying
• no mathematical operations possible (except counting)
• Ordinal (qualitative, but sometimes used quantitatively)
• ordered on some dimension
• Boolean operations possible
• Interval (quantitative)
• ordered with equal intervals
• addition, subtraction, and Boolean operations
• Ratio (quantitative)
• ordered, equal intervals, absolute and meaningful zero
• all mathematical operations possible

Problems with measurement of variables

• Qualitative vs. quantitative variables
• reliability and validity are essentially measurement problems
• since qualitative variables are basically classificatory, there is less concern with reliability and validity
• Reliability
• reliability refers to the observation of variation in scores earned by an individual on repeated trials of the same measure (variation can be systematic or random)
• so, reliability = consistency
• Validity
• validity is the degree to which the measuring instrument actually measures the concept in question
• validity also refers to the accuracy of the measurement
• it is possible to measure a concept more or less accurately if you are actually measuring the right concept but it is not possible to measure it accurately if you are not measuring it at all.
• Measurement error
• due to sampling
• due to subject or experimenter effects
• measurement error results in decreased reliability and validity

Relationship between variables

• X and Y are correlated if they vary together
• concomitant variation = correlation
• correlation can be direct or inverse

Causal relationships

• concomitant variation does not demonstrate causality
• causality is difficult (or impossible) to demonstrate logically
• However, we can make the case that X causes Y, if
• there is a relationship between X and Y(birds go south in the fall), and
• the relationship is asymmetrical so that a change in X results in a change in Y, but not vice versa (birds migrate because of fall but fall does not come because birds migrate), and
• a change in X results in a change in Y regardless of the actions of other variables, and
• generally, X should precede Y but sometimes symmetrical causality and simultaneous causality are allowed; the effect can never precede the cause
• Necessary and sufficient cause
• necessary ­ Y never occurs unless X also occurs (or has occurred)
• sufficient ­ Y occurs every time X occurs (but could also occur without X; e.g., "smoking causes cancer")
• necessary, but not sufficient (X must occur before Y but, X alone, is not enough for Y to occur; e.g., in order for me to shoot you with this gun (Y), I must point it at you (X), but X is not sufficient for Y)
• sufficient, but not necessary (X is sufficient to cause Y, but Z can also cause Y; e.g., Fred is wet (Y) but did he fall into a pond (X) or did he get caught in the rain (Z)?
• necessary and sufficient ­ Y will never occur without X and will always occur with X (e.g., the hand grenade will never explode without you pulling the pin and will always explode when you pull the pin)
• Causality in social science
• difficult to demonstrate theoretically as our theories are inadequate for the isolation of causes
• difficult to demonstrate methodologically
• survey methods usually do not give temporal sequences
• laboratory methods help to demonstrate causality since we control and sequence independent and dependent variables