Notes on Research in the Behavioral Sciences (Transcript)

Limits on Human Intuition

  • Several factors make it difficult for people to test theories about the world in an accurate manner.
    • The Overconfidence Effect
    • The Confirmation Bias
    • The Hindsight Bias
  • How much there actually is to know
  • How much you think there is to know
  • How much you know

The scientific attitude

  • How does the science of psychology correct for the limitations of human intuition?
    • 1) Applies amiable skepticism
    • When presented with information as fact:
      • Clarify terms (including success/failure of an idea)
      • How do you know that?
      • How good is the evidence?
      • Is there any evidence to the contrary?
      • What is the weight of the evidence?
      • Requires humility
    • Accept ideas supported by evidence
    • Remain skeptical of, but don’t completely discount, those lacking evidence
    • 2) Applies the scientific method

The scientific method

  • Start with a theory – General statement about how variables are related to one another
    • falsifiable
    • parsimonious
  • Develop a study and state your hypothesis – Derived from theory, more specific and testable
    • An a priori prediction of what will happen in the study
  • Analyze your data – Use statistics to evaluate your hypothesis
  • Report the results – Peer review process
  • Start all over again… A à B Except in cases of C, D, E, F, G H, I…

Using the Scientific Method in Psychology

  • We need to identify the study design(s) that best address our question(s)
    • Correlational research: observe & measure variables as they naturally happen
    • Cross-sectional – at one time point
    • Longitudinal – at multiple time points
    • Experimental research: Manipulate one variable and measure its effect on another
  • Why does design matter?

Cross-sectional research

  • Measure 2+ variables at one point in time in a large sample of participants
  • Variable B
  • Variable A
  • Represented by a correlation coefficient

Correlation Coefficients (r)

  • Describes relationship between two measured variables
  • Range: r∈[−1,1]r \in [-1, 1]
  • Positive correlations: 0 < r \le +1
    • Example: GPA and Attendance
  • Inverse (negative) correlations: -1 \le r < 0
    • Example: GPA and alcohol consumption
  • No correlation: r=0r = 0
    • Example: GPA and hair length
  • Interpretation scale (typical):
    • -1.0: Perfect negative
    • -0.8: Strong negative
    • -0.5: Moderate negative
    • -0.2: Weak negative
    • 0.0: No correlation
    • +0.2: Weak positive
    • +0.5: Moderate positive
    • +0.8: Strong positive
    • +1.0: Perfect positive

If you remember nothing else about correlational research, remember this

  • Correlation does not equal Causation

Let’s look into that further

  • Three ways to interpret a correlation:
    • A → B: Playing violent video games increases aggression
    • B → A: Aggressive people tend to like violent games more
    • C → A & B: A spurious third variable causes both
  • 3rd variable problem
  • Before we move along…
    • Correlations are not bad science, they’re just limited
  • Sometimes the only way to (ethically) study a topic
  • Directionality problem

3 Requirements of Causality

  • Causation requires:
    • Covariation (the variables vary together in a systematic way)
    • Temporal Precedence (the cause precedes the effect in time)
    • Elimination of Spuriousness (Ruling out alternative explanations)
  • Experiments
  • Longitudinal designs
  • Cross-Sectional designs
  • Longitudinal Experiments

SSS (Slide content illustrating an experimental design)

  • Independent Variable (IV)
  • Random Assignment
  • Dependent Variable (DV)
  • (Notes on the slide content in the transcript show some garbled text, but the core takeaway is the presence of an IV, random assignment, and a DV in an experimental layout.)

Sample vs Population

  • Goal: Describe a population
    • All cases in the group of interest
  • Population: The typical college student
  • Can we study all college students? Why not?
  • Sample: subset of that population that we use to represent the population
  • Random sampling: Each individual in the population has an equal chance of taking part in the study
  • Almost impossible with large populations
  • Convenience sampling: Sample of individuals easily available for study
  • Low likelihood of representativeness
  • Example: Study of drinking behavior of college students conducted at a religious university
  • Ways to correct for sampling bias
    • Restrict population
    • Sample in large numbers
    • Stratified sampling: Specifically recruit to find a representative sample of the population
    • Example recruitment composition: 59% non-Hispanic Caucasian, 19% Hispanic, 12.6% Black, 5% Asian, 1% Native American