Samenvatting aadvanced research methods

Advanced Research Methods

Knowledge clip 1

Dag theory
 Research question
o Must satisfy two conditions
 Open ended question
 Needs to include some causal language
 Dag theory
o Directed acyclic graphs are graphical representations of the
causal structure underlying a research question
 Exposure X to outcome Y





o Unbiased
 Correct for other factors such as diet and lifestyle (in this
case)
 Remove disruptive influence of other factors
o Dags
 Dags help to visualize the causal structure underlying a
research question
 You need a priori theoretical/subject knowledge about
the causal structure to draw a DAG
 E.g. from previous studies, literature, common
sense, etc.
 Collect data on all relevant variables
 Simle rules can be applied to determine for which
variables to add in regression analysis and how to
interpret the results
 Dag terminology
o Paths
 RQ: influence of x on y?
 A path is any route between exposure x and outcome y
 Paths do not have to follow the directions of the arrow

, 





o Causal paths and backdoor paths
 A causal path follows the direction of the arrows
 A backdoor path does not





o Open and closed paths, and colliders
 All paths are open, unless they collide somewhere on a
path
 A path is closed if arrows collide in one variable on that
path
 A variable in which the arrows collide = a collider

, 
o Blocking open paths
 Open (causal or backdoor) paths transmit association
 The association between x and y consists of the
combination of all open paths between them
 Here: all paths except x  w  Y
 To examine the influence of x on y, and only that, the
other associations that are not directly relevant need to
be removed > blocking those open paths
 Block those open paths  block open backdoor paths by
including a variable on that backdoor path in the
regression analysis
 An open path is blocked when we adjust for a variable L
along the …
 This means that we remove the disruptive influence of L
from the association between X and Y
 How? By including variable L in the regression analysis
 Backdoor paths always need to be closed
 Causal paths need to be open/closed depending on RQ

o Opening blocked path
 Include a collider in the analysis means you open the
blocked backdoor path
 Opening a backdoor path is disruptive for the association
between x and y  introduce bias of the influence
 Avoid bias by closing the backdoor path again  remove
the collider or add another variable that removes the
disruption

Lecture 1 causal inference
 Developments in quantitative research
o 1960-2010: focus on statistical methods
 Development of new techniques
 Improvements in computers and software
 Standardization of tests, focus on objectivity and
replicability

,  Helpful and harmful
o Since 2010: statistics is not so black and white
 Causal theory
 What (variables) should be part of a quantitative
analysis
 Statistics is not just about numbers
 Interpretation: meaning of results depends on context 
moving beyond p-value < 0.05
o Three different reasons for examining relationship between x
and y
 Description: patterns X and Y
 Prediction: Y given X
 Causal inference: Effect X on Y






o Causal inference
 Not interested in outcome per se
 Interested in the role of the treatment X on the outcome
 In an individual, a treatment has a causal effect if the
outcome under treatment 1 would be different from the
outcome under treatment 2
o Causal effect formal notation



