Summary Advanced Research Methods (Erasmus University Rotterdam: HCM)
46 views 1 purchase
Course
Advanced Research Methods
Institution
Erasmus Universiteit Rotterdam (EUR)
Complete summary of the course Advanced Research Methods at the Erasmus university Rotterdam (HCM/Zoma etc.)
I summarized all lectures and added relevant information from q&a's and workgroups throughout the summary.
Lecture 1 Advanced research methods (quantitative)............................................................................3
Lecture 2: Advanced research methods (quantitative).........................................................................15
Lecture 3: Advanced research methods (quantitative).........................................................................29
Lecture 4: Advanced research methods (qualitative)...........................................................................42
Lecture 5: Advanced research methods (qualitative)...........................................................................48
Lecture 6: Advanced research methods (quantitative & qualitative)...................................................54
,Lecture 1 Advanced research methods (quantitative)
Segment 1:
Assessing causal inference studies:
What was the question?
o What was the underlying question?
What was actually estimated?
o Is the estimate biased or unbiased?
o Is this an estimate of a full or partial effect?
Is the estimate really an answer to the question?
How was the analysis designed?
Were statistical methods applied correctly?
What is the estimate? Is that big, small, good, bad, etc?
o How uncertain is the estimate?
What do the researchers conclude? Is that conclusion justified?
Is this strong or weak evidence for something?
How does it compare with what we (thought we) knew?
Learning goals: You will be able...
To explain the potential outcomes approach in causal inference and apply it in thinking about
causal effect estimation.
To define ‘causal effects’
To apply the concepts of consistency, positivity, and exchangeability in randomised and non-
randomised settings.
Magazine ad:
- They are using causal language (e.g. ‘improves’ implies a causal effect)
o Causal effect: A leads to B
o using true match minerals leads to a better skin
- Is this study (from the ad) convincing:
o Problems in the study:
Small sample size
A small sample does not always have to be a problem. It
has a relation with the effect or outcome you are
measuring and the homogeneity. If for
example, all 41 women would drop dead
after using the product you are more likely
to believe it’s due to the product.
Study performed or financed by a commercial
company.
This is always a challenge, but you can do
things in order to protect your independence as a researcher.
Therefore, it is difficult but does not always have to be fatal.
No control group
This is actually an essential omission
They now don’t know what happens to women who do not use the
crème. (what happens without the treatment?)
, Potential regression towards the mean
o You can’t distinguish regression towards the mean from
actual effects of the product now. The women might also
have seen progression due to the fact that they use the
crème when their skin was bad and became better
throughout the days.
What do we want to know? (magazine ad)
- We are not interested in the outcome per se: ‘how many imperfections’
- We are interested in the role of treatment in achieving this outcome: ‘less imperfections
than without True Match Minerals’
- Conclusion: No meaningful causal conclusion can be drawn from this study
Causation (formal definition): ‘In an individual, a treatment has a causal effect if the outcome under
treatment 1 would be different from the outcome under treatment 2.’
- Woman A uses True match minerals: 2 spots
- Woman A does not use True match minerals: 5 bad spots
- Individual treatment effect: 3 spots (or 60%)
- Average treatment effect: average of individual effects in a population
The formula says ‘the outcome under treatment 1 is different than
the outcome under treatment 0.’
Potential outcomes:
- If user K uses the crème her skin will improve. If she doesn’t
use it her skin will not improve.
- If user L uses the crème her skin will improve. If she doesn’t
use it her skin will also improve.
- Etc.
The women however couldn’t be in both groups (they either used true
match minerals or they didn’t). Everyone got the crème. So: not all
potential outcomes are observed.
- Counterfactual outcome: potential outcome that is not
observed because the subject did not experience the
treatment. ‘It could have happened but it didn’t’.
- Potential outcome Ya=1 is factual for some subjects
and counterfactual for others.
The fundamental problem of causal inference:
, - Individual causal effects cannot be observed (think of the problem with counterfactual
outcome). You namely do not know what would have happened if the individual had gotten
no or the other treatment.
o Except under extremely strong (and generally unreasonable) assumptions.
- Average causal effect cannot be inferred from individual estimates. If you don’t have reliable
estimates of the treatment effect on an individual level, you cannot calculate the average
causal effect either.
o Causal inference can be seen as a problem of missing data!
- We need a different approach to causal effects. The solution are identifiability conditions (3).
Example:
- You go to the Lijnbaan in Rotterdam and ask everyone ‘Are you carrying a cigarette lighter?’.
After 20 years you come back to those people and ask them about their health.
o The causal question: ‘what is the effect of carrying a lighter on health?’
The average causal effects can still be determined under certain conditions.
- ‘Observing’ the counterfactual: what would have happened? (if the exposure had been
different)
- The 3 identifiability conditions apply to the question ‘what would have happened’.
- Based on population averages, causal effects can be estimated if three identifiability
conditions hold:
o Positivity:
Observe ‘what would have happened if…’
Positivity is about the sample and the way it was composed. Positivity means
that there has to be a positive probability for everybody in the sample of
being assigned to each of the treatment levels (you have a lighter or you
don’t).
You are either in one or the other group.
Units are assigned to all relevant ‘treatments’ within levels of adjustment
factors.
So you need both people with and people without cigarette lighters
in your sample.
People with a lighter could also not have had a lighter, and vice
versa.
In the example of L’Oreal this was not the case, because everyone
received the crème. Users could not not have used it.
You thus need a control group.
In the Lijnbaan example, the positivity seems to be okay.
o But because you now want to adjust for smoking status (see
exchangeability), we now need observations in 4 groups.
Smokers with & without lighters
Non-smokers with & without lighters
o Consistency:
Observe ‘what would have happened if…’
Define ‘if’: so give a clear definition of the treatment or exposure
, Example: ‘does water kill’ what do you mean with water? Are you
going to drink it? How much are you drinking? Is it rain? Are you
going to drown in it?
How do you specifically define your exposure (and its
counterfactual).
Is it consistent?
‘eat broccoli, it makes you healthy’: how do define healthy? What do
you mean by eating broccoli? (how much, how often, how is it
prepared etc.) And what are you comparing it to? (so how is not
eating broccoli defined?). So not consistent.
‘the effect of obesity on health’: how did they get obese? (eating a
lot of fries might have different effect on the body than a genetic
disorder). So again not consistent.
Effect of obesity on job prospects: if employers only look at people
to see how obese they are in their decision to hire them or not, it
might not matter how they (did not) get obese. In that context,
obese is a consistent concept.
Effect of healthcare spending on mortality: what is the money going
to be spent on? What is the alternative? How would the budget be
spent in the absence of the extra healthcare spending? Not very
consistent.
Carrying a lighter: this is consistent (you carry it in your hand or
pocket) and the control group is people who do not carry a lighter.)
o Exchangeability
Observe ‘what would have happened if…’
Treatment groups are exchangeable: it should not matter who gets
treatment A and who gets treatment B.
Notation:
‘The potential outcomes are independent of the treatment that was
actually received.’
Are people with and without lighters exchangeable (similar in other
respects)?
If you’d take all the lighters from the people who carry them and gave them
to those who didn’t carry them, would you get the same outcome (in
health)? The estimates would probably not be the same, because the carriers
probably smoke, which is a relevant difference between the groups that
should be taken into account.
You could take other factors into account (adjustment)
o You could look at only the group of smokers. are people
with and without a lighter in this group exchangeable?
stratification.
o Within non-smoking group, are people with and without a
lighter exchangeable?
If exchangeability is achieved once we zoom in on one of the groups, then
we can ascribe the association that we found to a causal effect.
If the three conditions are met, and we find an association in the data, then the association
is an unbiased estimate of the causal effect!
The benefits of buying summaries with Stuvia:
Guaranteed quality through customer reviews
Stuvia customers have reviewed more than 700,000 summaries. This how you know that you are buying the best documents.
Quick and easy check-out
You can quickly pay through credit card or Stuvia-credit for the summaries. There is no membership needed.
Focus on what matters
Your fellow students write the study notes themselves, which is why the documents are always reliable and up-to-date. This ensures you quickly get to the core!
Frequently asked questions
What do I get when I buy this document?
You get a PDF, available immediately after your purchase. The purchased document is accessible anytime, anywhere and indefinitely through your profile.
Satisfaction guarantee: how does it work?
Our satisfaction guarantee ensures that you always find a study document that suits you well. You fill out a form, and our customer service team takes care of the rest.
Who am I buying these notes from?
Stuvia is a marketplace, so you are not buying this document from us, but from seller danabechan. Stuvia facilitates payment to the seller.
Will I be stuck with a subscription?
No, you only buy these notes for $7.07. You're not tied to anything after your purchase.
