Objectives from Integrative Lit review

1
Objectives from Integrative Lit review

• Know what kinds of research questions can an
  integrative literature review answer.
• Describe 4 reasons meta-analyses might be lame
  and 3 reasons they might not be lame.
• Be aware of the judgment calls and possible
  effects of those judgment calls on the results of
  narrative and meta-analytic reviews (note the
  chapter is largely about the decisions reviewers
  make)

2
Interpreting Research Results

• February 28

3
Question

• Do we discover the truth or construct it?

4
Objectives

• Understand threats to valid inferences
• Distinguish odds ratio, relative risk,
  sensitivity, and specificity
• Consider chance as an alternative hypothesis
• Know possible sources of Type II error
• Describe the self-efficacy paradox, its relation
  to the changing signs of the SE-performance
  relationship, and its possible resolution

5
Threats to Valid Inferences

• Fallacy of the mean
• Group score does not apply to all in the group
• Omnibus test interpreted as directional and
  supportive
• Making ratio conclusions with less than ratio
  data
• E.g., attitude changed twice as much
• E.g., multiplicative element of expectancy X
  valence theories
• See below

6
Odds Ratio

• Index of strength of relationship (also called
  relative odds, which is not the same as relative
  risk)
• OR ad/bc
• RR a(b d)/b(a c) - If
  disease is rare, RR ? OR

7
Berksons Fallacy

• Danger is that these indices are sensitive to
  sampling bias.
• If sample is more likely diseased and exposure
  more likely to lead to disease, a is inflated
  relative to the population and hence odds ratio
  is inflated
• Is relative risk?

8
Sensitivity and Specificity

• Sensitivity is the ability to the
  predictor/screen/test to give a positive finding
  when the person is truly positive
• Specificity is the ability of the
  predictor/screen/test to give a negative finding
  when the person is truly negative

9
Screening
Validity
Truth
Sensitivity a/(a c)
a
Diseased
c
Not diseased
b
Specificity d/(b d)
d
Predictor(s)
10
Chance as an Alternative Hypothesis

• Validity is a question the interpretations of
  results
• Your interpretation(s) is(are) assessed against
  others
• Number of others with reasonable possibility
• Possibility of each
• If few, reasonable alternatives should be
  evaluated on their merits
• Statistics allows us to assess the probability
  that chance is an explanation if experimental
  hypothesis supported
• Or the probability that chance is not the
  explanation if null supported

11
Question

• Can/should we interpret a null finding as one of
  no relationship?

12
Possible Source of Type II
13
Other issues

• Stronger theories with point estimates (instead
  of simply, not zero)
• One persons bias is anothers theoretical
  perspective
• Presentation
• Sample presentation
• Example of interpretation issues
• Example of bias/theoretical perspective