Chapters 4

1
Chapters 4 5
2
Chapter 4 Studying Behavior

• Variables
• Independent controlled/measured by the
  investigator
• Dependent the observed outcome(s)
• Situational related to the environment/context
• Response kinds of behavior or performance
• Individual difference sets the person apart
  from the general human condition
• Mediating (or intervening) an intangible
  variable or one unobservable, such as hunger
  (eating can be observed but not hunger itself)

3
Operational definition of variables

• Many psychological processes cannot be directly
  observed (like hunger) but must be inferred from
  something which can be observed (like eating)
• An operational definition makes explicit exactly
  how you go about measuring or manipulating the
  variable, especially when it must be inferred
  rather than observed
• An operational definition can also specify the
  limits of what is being observed (like the
  specific acts counted as aggressive behavior) or
  the techniques used (e.g., WAIS for testing IQ)
• Operational definitions can also exclude unwanted
  influences or artifacts (e.g., if the person
  said they saw green we knew they misunderstood
  the instructions)
• Operational definitions are essential for
  replicability

4
Relationships between variables

• Two variables are related if observing one
  variable reduces your uncertainty about the other
  variable (mutual information) linear correlation
  is a special kind of mutual information
• Monotonic relationships go in one direction
• positive relationship the less you have of one
  variable then the less of the other, the more of
  the first one then the more of the other (e.g.,
  larger salary predicts larger bank balance)
• negative relationship the less you have of one
  variable the more of the other, the more of the
  first one the less of the other (e.g., lower
  expenses predict a larger bank balance)
• linear relationships imply proportionality
• Nonmonotonic relationships can be positive to a
  point but then negative (e.g., curvilinear,
  having an optimum range, or even more complex)
• E.g., how much you eat predicts life span eat
  too little and you are anorexic, eat too much and
  you are grossly obese, eat a more optimum amount
  to live longer

5
Determining relationships and causality

• Non-experimental methods
• Observe behavior as it occurs naturally, look for
  the things that covary
• What causes what (the chicken egg problem)?
• Did something else (mediating variable) influence
  both the things that co-varied?
• Experimental methods
• Observe behavior under controlled conditions
  (keeping as many extraneous variables as
  possible held constant)
• However, by definition this makes the situation
  artificial
• Some questions cannot be addressed at all
• Sometimes the artificiality hinders
  generalization to the natural setting
• Manipulating individual variables can shed light
  on causality and help to identify mediating
  variables, although this may be tedious
• Analysis techniques
• correlation, linear regression, etc., can
  establish a relationship
• Granger causality, partial directed coherence
  address cause

6
Is the research valid?

• Construct validity
• Did the operations and definitions accurately
  reflect the variable in question?
• Internal validity
• Were the logical arguments and operations carried
  out with enough precision to allow strong
  inferences?
• External validity
• Are the findings and/or inferences able to be
  generalized to people or situations not actually
  observed?
• See also face validity, statistical validity,
  etc.

7
Measurement Concepts

• Reliability (precision)
• Construct validity (accuracy)
• Reactivity (uncertainty principle)
• Kinds of measurement scales
• Nominal
• Ordinal
• Interval
• Ratio

8
Nominal measurement scale

• Like the numbers on football jerseys or your PIN,
  the number is equivalent to a name
• What does it mean to add, subtract, multiply, or
  divide with nominal scale numbers? (or calculate
  averages, etc.)
• Special statistics have to be used to analyze
  nominal scale measurements, such as Chi-square or
  binomial tests

9
Ordinal scale measurements

• Last week UF was 1, Miami was 2 this week
  Miami is 1, UF is 2. Is Miami twice as good as
  last week and UF only half as good? Is the
  difference between UF-Miami the same as the
  difference between 3 team and 4 team?
• What does it mean to add, subtract, multiply,
  divide, find average, etc.?
• Special statistics such as runs tests,
  Komolgorov-Smirnov tests

10
Interval scale measurements

• The difference between 1 and 2 is the same as the
  difference between 2 and 3, and so forth
• Typical example Fahrenheit thermometer
• Each degree F involves the same change of heat
• However, does zero deg F mean zero heat? (no)
• Is 100 deg F twice as much heat as 50? (no)
• You can get a meaningful result by adding or
  subtracting interval scale numbers, but not by
  multiplying, dividing, etc.
• Special statistics include randomization tests

11
Ratio scale measurements

• Like the ideal numbers we learned to add,
  subtract, multiply and divide in grade school.
  Zero means the stuff being measured is absent, 1
  and 2 differ by the same amount as 2 and 3, and
  also 2 means twice as much stuff as 1, 4 means
  twice as much stuff as 2, etc.
• Parametric statistics assume ratio scale
  measurements (t-tests, analysis of variance,
  etc.)
• Are any psychological variables being measured on
  a ratio scale (e.g., IQ)?

12
Special issue what is random?

• A variable whose value is influenced by a great
  many factors
• Thermal noise vs. pseudo-random noise (e.g.,
  random fractal with long-range self-similarity)
• Fair toss of a fair coin vs. unfair toss or coin
• Dripping faucet (deterministic chaotic process)
• Haphazard or without obvious systematic influence
• Humans are quite poor at generating randomness
• Gamblers fallacy
• Lets try it ourselves

13
Back to EXP 4174 Index
4174