Research and Statistics

1
Research and Statistics

• Introductory Psychology

2
Statistics

• 42.7 percent of all statistics are made up on the
  spot.

3
Stat Quotes

• There are three kinds of lies lies, damned
  lies, and statistics.
• Disraeli said it but attributed
  it to Mark Twain
• Statistics dont lie. There are liars, damned
  liars, and statisticians. Unknown
• You can lie with statistics, but you can lie
  better without them. Unknown

4
How Psychologists do Science

• Science can discover what ISNT true
• Scientific hypotheses can be disconfirmed
• Beliefs depend on evidence, observations
• Scientists agree on a set of standards for belief
• Scientists expose and give up erroneous ideas

5
Standards for Belief

• Falsifiability criterion
• Scientific evidence is public
• Evidence must be replicated
• Post hoc explanations dont count

6
Bad science example

• Natural cures they dont want you to know
  about, by Kevin Trudeau
• Currently 1 on NYTimes best seller list of
  how-to and advice books
• Contains NO references no one can check claims
• Claim A hospital in Mexico has virtually 100
  success rate in curing cancer with IV ozone
  hydrogen peroxide

7
Psychological Research

• Goal answer the question of cause. All human
  thoughts and actions are causes.
• Empiricism vs.. Rationalism Measurements
  expressed as data
• Psychological Experiment at least one variable
  (factor) is manipulated
• Quasi-experiments (lack random assignment)

8
Psychological Research cont.

• Non-experimental research methods include
  naturalistic observations, case studies, and
  correlational studies
• Independent variable (IV) the variable
  manipulated
• Dependent variable (DV) the outcome variable or
  measured variable
• Operational definitions

9
Does caffeine help memory?

• Independent variable?
• Dependent variable?
• E1
• E2
• C1
• C2

10
Does acupuncture help tension headaches?

• IV
• DV
• E
• C1
• C2

11
Does coaching for SAT work?

• Pre-SAT Coaching Post- SAT
• E 1100 3 mos 1200
• Would you say that coaching caused a 100 pt
  increase in performance on the SAT?

12
Does coaching for SAT work?

• Pre-SAT Coaching Post- SAT
• E 1100 3 mos 1200
• C 1100 ------- 1190
• How much improvement was due to coaching?

13
3 Essential Properties of a Well-Designed
Experiment

• The researcher must vary at least one IV to
  assess its effects on behavior
• The researcher must have the power to assign
  participants to the conditions in a way that
  assures their initial equivalence, e.g., randomly
  
• The researcher must control extraneous variables
  that may influence Ss behavior

14
Home school vs. public school

• H home schooled kids have higher SATs?
• IV type of school
• DV SAT scores
• E Home Schooled
• C Public Schooled
• What is wrong with this design?

15
Two uses of Random1. Random Selection

• Population the entire group to which we wish to
  generalize our findings
• Sample a subset of a population
• Random sample a method of selecting our sample
  such that all members of the population have an
  equal chance of being selected a representative
  sample, an unbiased sample

16
2. Random Assignment

• In experiments, which person gets which treatment
  (IV) is determined by random chance.
• Called random assignment to groups or
  conditions, e.g., real medicine or placebo
• Random is a method, not a result

17
Biases and Placebos

• Experimenter Bias E transmits expectations that
  influence results
• Subject Biases Subjects or patients beliefs
  or expectations influence the results, e.g.,
  demand characteristics and placebos
• Controls for biases single blind, double blind,
  triple blind

18
Cautions

• Just because you believe something is true,
  doesnt mean it is true
• There is a strong, emotionally driven tendency
  for people to believe what they want to believe
  and then to look for confirmatory evidence
  (confirmation bias) and discount or ignore
  contradictory evidence.

19
Cautions,cont.

• Anecdotal evidence is not scientific evidence
  (person who) arguments
• One of the worst outcome measures is asking a
  person if he or she feels better or if he or
  she believes something worked. Without proper
  controls, including placebo controls, one usually
  cannot determine if something worked.