What is p-hacking and how do you do it?
Running multiple versions of a model or selecting subsets of data until p < .05. Adding covariates, excluding participants, running more participants.
When is cross-sectional mediation okay?
I find differences in emotion regulation between patients with depression and healthy controls. Does that mean emotion regulation increases when people get less depressed?
Nope! Always make sure to describe differences (greater/more) in cross-sectional designs and save changes (increases/improvements) for longitudinal designs.
How would you study mediation vs. moderation in a longitudinal design?
Mediation: Pick variables you think change over time, measure them at multiple time points, test the product of the effect from x to m and m to y.
Moderation: Pick an individual difference variable and test the product of it with another variable.
Why does random assignment allow us to claim causality?
It ensures groups come from the same population and so are identical on all unmeasured characteristics.
When do you NOT need to cite a source?
Never! Always cite your sources (maybe not if it's a very general statement).
What is a p-value?
The probability of observing a value as large or larger under the assumption that there is truly no effect.
I run a regression in a cross-sectional design – can I say one variable predicts the other?
Yes and no. In a regression framework, one variable does predict the other, but it can also imply a causal inference that is unwarranted so it's best to say they're associated!
What do you need to include in an RCT to test mediational/process/mechanistic questions?
Why do (non-randomized) single case experimental designs allow us to claim causality?
Should you share your data? Why/why not?
Yes - promotes greater accessibility & larger sample sizes.
No - you could get scooped or data could be used for unintended purposes.
Why is it important to disaggregate between- from within-person effects?
Between-person effects characterize individual differences. Within-person effects characterize how a given person changes over time. These are different types of variability with different inferences (e.g., error rates in fast typers vs. typing faster than normal).
Are there any difficulties in assessing lifetime prevalence of a disorder in people at different ages?
Lifetime prevalence is more likely in people who have lived longer! (Although if they've lived longer, they may also be healthier)
What’s the difference between mediators and mechanisms?
Mediators: Statistical constructs (product of x to m and m to y)
Mechanisms: Conceptual constructs (mediators + theory/experimental evidence)
Why can it be important to run experiments on psychopathological processes using people without psychopathology?
To test how a construct leads to the development of psychopathology, rather than correlates with existing aspects of psychopathology.
Aren't all hypotheses conflicts of interest? Why/why not?
Yes - it makes you invested in a particular outcome.
No - I preregistered my study to limit this conflict!
Name 2 types of internal validity and 2 types of external validity.
Internal: internal consistency, test-retest reliability, construct validity, content validity, criterion validity, face validity, convergent vs discriminant validity, 3rd variable confounding, selection bias, differential attrition
External: population validity, ecological validity
How much within-person variability is present in cross-sectional designs?
0
What are the 3 types of prevention trials and their general sample sizes?
1. Universal prevention (30,000 - 3 million)
2. Selective prevention (1,000 - 2,000 or fewer)
3. Indicated prevention (30 - 100)
Give me one example of a treatment developed from experimental work. (Bonus pts if you know how efficacious it is)
Cognitive bias modification - not that efficacious!
How do you decide authorship on a team?
Group discussion before beginning to outline roles, scope, & authorship. Clearly defining contingencies. Checking with mentor to ensure your bases are covered.
What's the difference between statistical significance, effect sizes, and clinical significance?
Statistical significance: Is observed value as large or larger than we would expect assuming no effect?
Effect size: Dividing an effect by a standardized value to create a unit-less measure of effect strength.
Clinical significance: Do patients report a noticeable change?
What is Berkson’s bias/Lord’s paradox/Simpson’s paradox? How does it affect inferences?
The idea that dividing groups based on a cut-off score that is not perfectly reliable will lead to weaker or spurious negative relations among variables.
What are the 5 stages of translational science?
0. Basic
I. Creating/modifying/adapting intervention
II. Testing efficacy in research setting
III. Testing efficacy in community setting
IV. Testing effectiveness in community setting
V. Dissemination & Implementation
Is correlation necessary to (eventually) claim causality?