Chapter 13: Interpreting Research Results

1
Chapter 13 Interpreting Research Results

• Describing Results
• Inferences in Behavioral Science Research
• Null Results
• Integrating Results of Research
• Summary

2
Describing Results

• Nature of relationships
• Types of Relationships
• Linear v. Curvilinear
• Mediators and Moderators (partial corr or MR)
• Interaction (factorial experiments)
• Predicted and Observed Relationships
• Cf results (observed) to expected (hypothesis)
• Table 13-1, p. 428, Table 13-4, p. 429

3

• Real v. Chance Relationships
• Inferential Stats (what alpha level? Why p lt.05?)
• Type I and Type II error trade offs
• Testing the Proper Statistical Hypothesis
• Multiple tests (what effect on alpha level?)
• Omnibus (MANOVA) v. Planned comparisons
• Whats the benefit of Planned comparisons?
• Effect Size and Importance of Effect Size
• Effect size (always include effect size)
• http//web.uccs.edu/lbecker/Psy590/es.htmII.20in
  dependent
• Pearson r Cohens d
• Practical Significance (small, medium, large?)

4
Effect size

• When effect size is small)
• -gt Weak manipulation -gt any effect may be
  important
• -gt When everything else is equal, it may play an
  important role
• ? What are the practical implications for the
  small effect sizes found in
• Tajfel et al., 71 minimal group effect and
• Hatfield Sprecther, 86 physical
  attractiveness?

5
Practical SignificanceSmall effect sizes

• Clinical significance (a value judgment)
• Abelson (85) skill and batting average (r .06)
• Important over a whole season
• Fishbein Ajzen (75) religiosity and religious
  behavior
• Small effect size and large populations
• Framington study (Rosenthal Rosnow, 91)
• Asprin and avoid heart attack (r .03)
• Population of 750k people decrease of 3.4
  heart attack rate
• Theory testing v. Applied research
• Which is effect size more important for? (Chow,
  88)
• Applied research

6
Inference in Behavioral Science Research

• Knowledge as a Social Construction
• Constructionist viewpoint
• Do we build our own reality? Or
• Is logical positivism a real possibility?
• ?How do we view the cause of racial prejudice
  now?
• What zeitgeist are we in now? (Table 13.3, p 436)
• Blank slate? Or biological evolution (cognitive)?
• Bias in Interpreting Data
• Theoretical bias (e.g. Mony Ehrhardt, 72)
• Which interpretation is correct?
• E.O. Wilson (78) sociobiologist or
• Mackie (83) cultural influence to explain
  results

7
Inferences Bias

• Personal Bias (tenacity)
• Sherwood Nataupsky (68) study of 82
  psychologists beliefs about racial differences
  in IQ
• Environmentalists
• Hereditarians
• Middle-of-the-roaders (inconclusive)
• Statistical sig differences (Bias shows up)
• Larry Summers (What happened to him? Why?)
• Assuming group differences are biological /
  environmental
• Correlational data make it hard to decide
• Victim blame (look beyond the group for theory)
• Behavior labeling (aggressive v. assertive)

8
InferencesMaking Valid Ones

• Measurement and Statistics
• Know the level of measure
• Recognize the fallacy of the mean
• E.g. distributions overlap
• State correlational results and group means
  appropriately
• Corr state direction and strength
• E.g. positively related
• high scores on X were associated with high
  scores on Y
• Group means
• mean for group A was significantly higher than
  the mean for group B
• Dont forget to show group means (ANOVA table
  doesnt)
• ? State your hypothesis(es) without falling prey
  to the fallacy of the mean

9
Valid Inferences

• Empiricism
• Stay close to the actual statistical findings,
  dont speculate until the discussion
• Clarify (or qualify) the relationship between the
  hypothetical construct and op definition
• E.g. how is race (hypothetical construct) defined
  operationally?
• Describe, avoid unwarranted evaluations
• E.g. do women underestimate the credit they
  deserve or do men overestimate? (you know the
  truth!)
• Causality
• Dont infer causality from correlational findings
• Generalization
• Theory and or findings

10
Inferences3 Uses of the Null Prejudice
against Null

• Testing hypotheses
• Research validity
• Testing generalizability
• Null findings dont get published (despite the
  fact they may be well done)
• If the null is in fact true, What does this imply
  about the published studies?
• They may be Type I errors!
• ? Researchers unlikely to test the null directly
• Why?

11
Possible Sources of type II Errors

• IV
• Construct valid?
• Manipulation effective?
• strong enough?
• DV
• Construct valid?
• Sensitive enough?
• Unrestricted range?
• Design
• Curvilinear relationship? (inspect the
  distribution)
• Extraneous vars controlled?
• Moderators or mediators operating?
• Large enough sample (power test)

12
Accepting the Null

• Common criteria
• Proper design and Sufficient power
• Predicted null results
• Based on good theory
• Unexpected null results
• Theory could be wrong! (believe it or not)
• Suppose it is a Type II error?
• Cold Fusion Another chance. Does theory matter?
  Cost of Type II error
• ? What could explain why you may not find support
  for your hypotheses?

13
Integrating Results

• Identifying Implications for Theory
• Comparison with prior research
• Comparison with theoretical prediction
• Identifying Implications for Research
• Research procedures
• New research questions
• Identifying Implications for Application

14
Chapter 13 Interpreting Research ResultsSummary

• Describing Results
• Inferences in Behavioral Science Research
• Null Results
• Integrating Results of Research
• Summary