PSYCH 242 FINAL EXAM STUDY GUIDE
242-20 Evaluating Bivariate Association Claims (Ch. 8: p. 210-230) - Answers - 242-20
Evaluating Bivariate Association Claims (Ch. 8: p. 210-230)
Discuss how the four big validities apply to association claims (i.e., claims based on
correlational research) and which have higher priority.
[pg.210] - Answers - • Construct validity (high priority): How well was each variable
measured?
• Statistical validity (high priority): How well do the data support the conclusion of
association?
• External validity (somewhat important): To whom and what can the association be
generalized?
• Internal validity (not usually relevant): Can we make a causal inference from
association? (No)
Be able to interrogate the reliability and construct validity of measures in an association
claim.
[pg.211] - Answers - Reliability: How to test it
• Test-retest reliability for measures of stable variables
• Inter-rater reliability for non-automated observational measures
• Internal reliability for multi-item self-report measures
Construct Validity:
• Does the measure have face validity for the variable?
• Does it tap into all aspects of the variable (content validity)?
• Is it associated with predicted quantitative or categorical outcomes (criterion validity)?
• Does it correlate more strongly with established measures of the same variable
(convergent validity) than with measures of other variables (discriminant validity)?
List and describe elements of statistical validity in association claims: effect size,
prediction accuracy, and importance; statistical significance; and how outliers, restricted
range, or curvilinear relationships may result in false positive (Type I) or false negative
(Type II) errors.
[pg.212-] - Answers - Effect Size [pg.210-11]: describes the strength of an association;
Strong associations = strong effect sizes
-stronger effect sizes permit more accurate predictions
Prediction Accuracy [pg.212]:
-Larger effect sizes give more accurate predictions
-The more strongly correlated 2 variables are (the larger the effect size), the more
accurate our predictions can be.
Importance [pg.213-14]: stronger effect sizes typically more important
, Statistical Significance [pg.214-15]: refers to the probability of obtaining an association
of that strength by chance, assuming that there is actually no association in the
population
• This probability is the p-value; p < .05= significant
-p-value is dependent upon both effect size (large effect size=statistically significant)
and sample size
Outliers [pg.217]:
-can result in false positives or negatives
-they skew the data b/c their value is much higher or lower than the norm
-in bivariate correlations, outliers are problematic when they involve extreme scores on
both variables
- have more impact when the sample is small
Restricted Range [pg.218]:
-Not incorporating all values of obtained data; in a correlational study, if there is not a
full range of scores on one of the variables in the association
-might make a sample's correlation appear smaller than it really is in the population
Ex: only using people who scored 1800 or more on SAT, in a college success study
Curvilinear Relationships [pg.220-1]:
-An "inverted-U" or U-shaped curve
-Pearson r will yield a small, nonsignificant value in this case (may be a false negative)
Type 1 error (false positive):
Type 2 error (false negative):
Be able to interrogate the external validity of a correlational research study and assess
the importance of a representative sample and generalizability of an association claim. -
Answers - • Does the association generalize to other people, places,
and times?
• How the sample was selected from the population of interest is what matters for
external validity, not the sample size
•A lack of external validity should not disqualify an entire study. If the study fulfills the
other 3 validities & has sound results, the question of generalizability can be left for
further investigation
Explain two reasons why it is invalid to draw causal conclusions from correlational
research (directionality problem and third-variable problem), and be able to identify and
explain reasonable alternate explanations for a given correlational finding. - Answers - -
Directionality problem: (no temporal precedence)
• Even if causal, not always possible to know the direction of causation (A may cause B,
or B may cause A, or reciprocal)
242-20 Evaluating Bivariate Association Claims (Ch. 8: p. 210-230) - Answers - 242-20
Evaluating Bivariate Association Claims (Ch. 8: p. 210-230)
Discuss how the four big validities apply to association claims (i.e., claims based on
correlational research) and which have higher priority.
[pg.210] - Answers - • Construct validity (high priority): How well was each variable
measured?
• Statistical validity (high priority): How well do the data support the conclusion of
association?
• External validity (somewhat important): To whom and what can the association be
generalized?
• Internal validity (not usually relevant): Can we make a causal inference from
association? (No)
Be able to interrogate the reliability and construct validity of measures in an association
claim.
[pg.211] - Answers - Reliability: How to test it
• Test-retest reliability for measures of stable variables
• Inter-rater reliability for non-automated observational measures
• Internal reliability for multi-item self-report measures
Construct Validity:
• Does the measure have face validity for the variable?
• Does it tap into all aspects of the variable (content validity)?
• Is it associated with predicted quantitative or categorical outcomes (criterion validity)?
• Does it correlate more strongly with established measures of the same variable
(convergent validity) than with measures of other variables (discriminant validity)?
List and describe elements of statistical validity in association claims: effect size,
prediction accuracy, and importance; statistical significance; and how outliers, restricted
range, or curvilinear relationships may result in false positive (Type I) or false negative
(Type II) errors.
[pg.212-] - Answers - Effect Size [pg.210-11]: describes the strength of an association;
Strong associations = strong effect sizes
-stronger effect sizes permit more accurate predictions
Prediction Accuracy [pg.212]:
-Larger effect sizes give more accurate predictions
-The more strongly correlated 2 variables are (the larger the effect size), the more
accurate our predictions can be.
Importance [pg.213-14]: stronger effect sizes typically more important
, Statistical Significance [pg.214-15]: refers to the probability of obtaining an association
of that strength by chance, assuming that there is actually no association in the
population
• This probability is the p-value; p < .05= significant
-p-value is dependent upon both effect size (large effect size=statistically significant)
and sample size
Outliers [pg.217]:
-can result in false positives or negatives
-they skew the data b/c their value is much higher or lower than the norm
-in bivariate correlations, outliers are problematic when they involve extreme scores on
both variables
- have more impact when the sample is small
Restricted Range [pg.218]:
-Not incorporating all values of obtained data; in a correlational study, if there is not a
full range of scores on one of the variables in the association
-might make a sample's correlation appear smaller than it really is in the population
Ex: only using people who scored 1800 or more on SAT, in a college success study
Curvilinear Relationships [pg.220-1]:
-An "inverted-U" or U-shaped curve
-Pearson r will yield a small, nonsignificant value in this case (may be a false negative)
Type 1 error (false positive):
Type 2 error (false negative):
Be able to interrogate the external validity of a correlational research study and assess
the importance of a representative sample and generalizability of an association claim. -
Answers - • Does the association generalize to other people, places,
and times?
• How the sample was selected from the population of interest is what matters for
external validity, not the sample size
•A lack of external validity should not disqualify an entire study. If the study fulfills the
other 3 validities & has sound results, the question of generalizability can be left for
further investigation
Explain two reasons why it is invalid to draw causal conclusions from correlational
research (directionality problem and third-variable problem), and be able to identify and
explain reasonable alternate explanations for a given correlational finding. - Answers - -
Directionality problem: (no temporal precedence)
• Even if causal, not always possible to know the direction of causation (A may cause B,
or B may cause A, or reciprocal)
