Week 2: Study Design Flashcards
Correlational studies
Corelates are variables that are associated at a given point in time
It is not clear that one predates the other
(1) Natural experiments or case-control designs
Compare kids with disorder to kids without
Have to ensure that groups are matched on other variables
eg rates of depression in kids who have and have not experienced abuse
Must be vary careful about controlling for other variables because it affects internal validity
Internal Validity
The extent to which group differences are due to the variable of interest
Case-control designs advantages/disadvantages
Allows one to answer important questions such as why a group is different
Often it is the only ethical way to address a question - you cannot assign people to psychopathology!
Helps to rule out other explanations
You can establish internal validity (if you account for confounding variables)
WEAKNESSES
Does not establish temporal order
Does not establish that A changes B (responsiveness)
Prospective Designs
Identify children and follow them over time
You can get a temporal link with this
You could establish what factors put children at risk for something
You can account for what happens to children who have a disorder - allows predictions
Risk and protective factors occur before the children get the disorder and either increase or decrease the risk. If you are monitoring children you will see these play out.
eg child abuse and conduct disorder
high-quality relationships with supportive adults and depression
You can then track the children with a disorder and find out what happens to them
eg are psych symptoms stable over time
does psych in childhood correlate with more global problems in impairment (social issues or employment difficulties etc)
Prospective Designs Advantages/disadvantages
ADVANTAGES
Establishes temporal order
Not relying on possibly flawed, subjective retrospective reports
DISADVANTAGES
Takes a LONG time
Internal validity: You must be very careful to rule out the effects of other factors
Especially age as sometimes things change just via development (so is it the variable of interest that changes with the DV or is the change in DV a function of aging?)
Could be very hard to find enough people if the condition is rare (and base rate is low).
eg autistic children that get autism super young. It is advantageous for the child to get treatment asap so would be good to study this but it is super rare so you would need a massive study to find many of them.
Does not establish that A changes B
Best practise guidelines
Are based on scientific approaches
Allows us to be confident that the treatment improves symptoms and reduces impairment.
These come about in one of two ways:
(1) Based on research findings which identify treatment for which there is scientific evidence of efficacy
However science is slow and many gaps in the literature remains so in this case we use
(2) expert-consensus approach wherein a panel of experts makes a ruling which fills a gap in the literature until such a time as research fills it.
A well established treatment has (one of two things)
(1) A large series (>+9) single-case study designs demonstrating efficacy
OR
(2) At least 2 between-group design experiments
(this might be changing towards systematic review of the literature and a committee reviewing the evidence)
Single-Case experimental designs
A-B-A-B
Examine the effect of a treatment on a single child’s behavior
This requires multiple measures of the behavior and the treatment effects.
A-B-A-B Reversal design A - baseline of behavior B- intervention phase A - return to baseline (remove intervention) B - reintroduce intervention
eg, kid gets up too much at school. Give him 7 sticks, take one away when he gets up, if he has any left at the end of the day he gets a prize. Observe behavior Sttop this. Observe change in behavior Start again Observe behavior
A researcher would want to see improvement when the intervention starts (better than baseline), then a deterioration when intervention ends (return to baseline) then an improvement when the intervention is reintroduced.
In reality this will be noisy with ups and downs but overall the trend should go this way.
Single-Case experimental designs (Adv/Disadv)
ADVANTAGES
Internal validity: The external event happened at the same time as the onset of the intervention.
Stopped as it stopped.
Very unlikley that this timing would be down to total chance.
Temporal ordering
A changes B (Responsiveness)
Possible to infer causality
DISADVANTAGES
External validity: You are only working with one child. How generalizable is this? This is why there must be 9 of these.
Can be hard to interpret the findings (especially if very noisy with many improvements and deteriorations)
Stable change : The intervention might be so effective that the behavior never goes back to baseline. This is bad for researchers but great for clinicians.
Ethics: If you find an intervention that works for a child against a debilitating condition, it is totally irresponsible to remove it (and you would not because this would be unethical). In this way, it is hard to use this design to show it works.
Random Control Trials (RCTs)
Use random assortment so other non IV characteristics should balance out.
Have an experiment and control group
Compare the results
RCTs: Internal and Construct validity
Internal validity: Is it my intervention that is causing the change in outcomes?
Helped by random assortment.
Construct validity : What about my intervention is causing the change in outcome?
RCTs: Control Groups
If you only compare your intervention against nothing, you can only say it is better than nothing. Not necessarily that your specific intervention worked.
There are four ways to make a control group for an RTC in therapy:
(1) Wait list condition
They get nothing
But they will get the intervention afterwards
(2) Attention-only control group
The control group meet a therapist who provides UPR but nothing else
(3) Treatment as usual
The control group participant gets whatever they would have done if they were not in the trial
(4) Another, effective treatment
Compare to a validated treatment
RCTs: Advantages/Disadvantages
ADVANTAGES
Internal validity: very good for knowing our intervention accounts for the changes
Construct validity: with a carefully chosen control group is also very good
Causality
DISADVANTAGES
External validity: We need to know if it works with other people, does not always show this because of sampling which is often WEIRD and not representative of real-world clinical populations (eg working class people). Comorbidities are often excluded so that researchers can be sure it is effective against the target condition but this does not represent real clinical populations well where comorbidity is common.
Context: often done at universities doing detailed, manualized treatment often by very well-trained therapists (often the graduate students of the person who developed the treatment allowing for huge training and supervision).
Drop out: Participants do not like being randomly assigned and put in control groups. They often guess that this has happened and then figure this is not helping them so drop out. This threatens internal validity as the people that stay might differ from those who drop out, leading to the control and treatment group differing in significant ways.
RTCs use averages: if on average there was improvement, some people likely did not. Some will have gotten worse. Now people try to work out why so they may predict for whom the treatment works.
RTC summary
Powerful test of intervention efficacy (because of random assignment)
Powerful test of theory about psychopathology - you should be able to target a part of the condition as it is theorized. The effectiveness of this intervention also tests the underlying theory.
Can establish causality
And responsiveness, can show A changes B
