Tentamen vragen en uitwerkingen Evidence-Based Clinical Practice
Q1
Given information question 1
A researcher investigates the effects of antidepressant (vs. placebo) on a post-treatment depression score in seniors.
She wants to know if the effect of antidepressants depends on loneliness, as assessed (on a continuous scale) before
the intervention.
Questions
1) Why can’t this question be addressed with the correlation analysis? Please elaborate briefly.
2) What type of analysis is appropriate to address this question? Please elaborate briefly.
1) The key here is an interaction effect: is the effect of intervention (drug vs. placebo) moderated by loneliness?
Correlation analysis examines the association between two variables; but cannot expose interaction effects.
(Still, you could correlate loneliness to the difference in the depression score before and after treatment, but
before treatment was not measured).
2) Moderation analysis does allow testing such a dependency. It allows to test a specific interaction: whether the
effect of interventions depend on some moderator variable.
- Note that in ANOVA with loneliness as a covariate, the interaction effect between interventions (IV) and
loneliness (CV) gives the exact same answer as a moderation regression analysis. They still should mention
the term moderation though.
Own answers question 1
1) The researcher wants to check if the feeling of loneliness might influence the outcome while giving the seniors
either antidepressants or placebo. So in other words, it is the question if loneliness interacts with the sort
treatment where an interaction effect between loneliness and the intervention is needed to be analyzed. An
correlation analysis only looks at the possible difference between the outcome of the two groups, but not at
this interaction. Therefore you need an:
2) Moderation analysis is appropriate here because you want to establish if the effectiveness of the treatment
depends on the loneliness of the seniors. The variable ‘loneliness’ is an continuous moderator that itself
independent of getting either an antidepressant of placebo treatment and therefore can possible affect the
effect of the treatment. If the interaction of loneliness has an significant interaction effect with the sort
treatment someone receives, then loneliness was a moderator of the effects of treatment on the depression
score in seniors.
Theory behind it
Moderation:
Sex (nominal):
run a regular ANOVA; intervention: yes/no and sex (m/v). If there is an interaction, then there is an moderator
of the effects of treatment on wellbeing.
Intelligence (continuous). The treatment depends on variables that are themselves independent of treatment.
Run a ANCOVA with intervention: yes/no and IQ as an covariate. If there is an interaction, then indeed IQ is a
moderator of the effects of treatment on wellbeing.
, Q2
Given information question 2
One wants to know if the present volume of certain brain areas, namely the hippocampus and the prefrontal cortex, is
a good predictor for dementia 5 years from now. A logistic regression analysis is performed in which the model
describes how the risk of dementia depends on the volume.
Questions
1) Why is logistic regression the designated analysis technique?
2) Based on the SPSS output, is the volume of the hippocampus a good predictor of dementia, and if so, what is
the direction based on the test results below (max 2 lines).
3) Why is the scatterplot with prefrontal cortex volume on the x-ass (see below) so ‘messy’?
Answers question 2
1) The dependent variable is discrete (binominal).
2) Yes, if the term is omitted, then the model fit significantly deteriorates ( p =.018). The volume increases, the
chance of dementia decreases.
3) Because prefrontal volume is not a significant predictor of dementia.
Own Answers question 2
1) The question: Does the risk of dementia depends on the volume of certain brain areas? Where you can
Q1
Given information question 1
A researcher investigates the effects of antidepressant (vs. placebo) on a post-treatment depression score in seniors.
She wants to know if the effect of antidepressants depends on loneliness, as assessed (on a continuous scale) before
the intervention.
Questions
1) Why can’t this question be addressed with the correlation analysis? Please elaborate briefly.
2) What type of analysis is appropriate to address this question? Please elaborate briefly.
1) The key here is an interaction effect: is the effect of intervention (drug vs. placebo) moderated by loneliness?
Correlation analysis examines the association between two variables; but cannot expose interaction effects.
(Still, you could correlate loneliness to the difference in the depression score before and after treatment, but
before treatment was not measured).
2) Moderation analysis does allow testing such a dependency. It allows to test a specific interaction: whether the
effect of interventions depend on some moderator variable.
- Note that in ANOVA with loneliness as a covariate, the interaction effect between interventions (IV) and
loneliness (CV) gives the exact same answer as a moderation regression analysis. They still should mention
the term moderation though.
Own answers question 1
1) The researcher wants to check if the feeling of loneliness might influence the outcome while giving the seniors
either antidepressants or placebo. So in other words, it is the question if loneliness interacts with the sort
treatment where an interaction effect between loneliness and the intervention is needed to be analyzed. An
correlation analysis only looks at the possible difference between the outcome of the two groups, but not at
this interaction. Therefore you need an:
2) Moderation analysis is appropriate here because you want to establish if the effectiveness of the treatment
depends on the loneliness of the seniors. The variable ‘loneliness’ is an continuous moderator that itself
independent of getting either an antidepressant of placebo treatment and therefore can possible affect the
effect of the treatment. If the interaction of loneliness has an significant interaction effect with the sort
treatment someone receives, then loneliness was a moderator of the effects of treatment on the depression
score in seniors.
Theory behind it
Moderation:
Sex (nominal):
run a regular ANOVA; intervention: yes/no and sex (m/v). If there is an interaction, then there is an moderator
of the effects of treatment on wellbeing.
Intelligence (continuous). The treatment depends on variables that are themselves independent of treatment.
Run a ANCOVA with intervention: yes/no and IQ as an covariate. If there is an interaction, then indeed IQ is a
moderator of the effects of treatment on wellbeing.
, Q2
Given information question 2
One wants to know if the present volume of certain brain areas, namely the hippocampus and the prefrontal cortex, is
a good predictor for dementia 5 years from now. A logistic regression analysis is performed in which the model
describes how the risk of dementia depends on the volume.
Questions
1) Why is logistic regression the designated analysis technique?
2) Based on the SPSS output, is the volume of the hippocampus a good predictor of dementia, and if so, what is
the direction based on the test results below (max 2 lines).
3) Why is the scatterplot with prefrontal cortex volume on the x-ass (see below) so ‘messy’?
Answers question 2
1) The dependent variable is discrete (binominal).
2) Yes, if the term is omitted, then the model fit significantly deteriorates ( p =.018). The volume increases, the
chance of dementia decreases.
3) Because prefrontal volume is not a significant predictor of dementia.
Own Answers question 2
1) The question: Does the risk of dementia depends on the volume of certain brain areas? Where you can
