- Introduction to causal inference Read
- Causal impact, d-separation and backdoors
- Blocking backdoor
- Example of back door criterion
- Minimizing stratification through backdoor criterion
- Network analysis using Grow Shrink & Hiton & Sequence
R code► Slides►
- Network analysis using Poisson regression
- Optimizing stratification Read►
- Impact of sequence on accuracy of network learning algorithms
For this assignment you can use any statistical package. Work can be done in group's of two students but you cannot work with a student that you have previously teamed up with.
Question 1: Inside an electronic health record, there are data
on outcomes of a particular intervention. Using the network drawn
below, write the equations that would allow you to estimate what would
happen if the intervention was not given. First, write an equation
for each node in the network based on variables that precede it.
For example, the regression equation for predicting whether there is an
adverse event is given by the equation:
Outcome = a + b Treatment + c Severity
Second, set the variables that change across these
equations to the relevant values. For example, set Treatment to be
Question 2: The following graph was used to simulate data on bundling
payment for total hip fracture
Recover the original network and calculate the causal
impact of H on BP.
Detail R Code►
If you were using logistic or ordinary regression
equations, write what set of equations are represented by the above
network. In each instance write all the variables that are in
the regression equation and the variables that have a statistically
signficant relationship with the response variable. For example,
LTH is regressed on all variables that precede it which are DME, CL, P
and H. But only P and H have a statistically signficant
relationship with LTH. This regression can be shown as:
LTH = a + b DME + c CL + d P* + e H*
In the above equation, the
statistically signficant relationships are shown with a star (*).
A missing star indicates an insignificant relation. Using the
data, estimates the parameters of each of the regressions. Can
these set of equations be used to create the network. In how
many ways does the regression equations differ from the network model
in the graph.
Question 3: Construct a model to predict what would have happened if patients
who received citalopram in level 1 of the STAR*D experiment had received
a different medication.
- Read about the STAR*D study protocol.
- Download data. Use instructor's last name as password.
- Create data sets for each antidepressant combination (variable
Concat). This data sets will include a patient several times,
if the patient received different combinations of antidepressants
over time. Group By Concat and ID variable to
remove the weekly data.
- Identify the parents in the Markov Blanket of each
antidepressant. You can use ordinary regression or logistic
regression to do this.
- Stratify treatment and exclude from the list of parents in the
Markov Blanket any variable not related to remission (measured as
referred to followup).
- Calculate the unconfounded impact of antidepressant(s) on
remission. Stratify the remaining variables in the parents in
the Markov Blanket of treatment and calculate the impact of
antidepressant(s) on remission.
- Repeat the analysis for at least 3 antidepressant(s).
- Evaluate for a patient with PTSD and neurological disorders
which antidepressant is most likely to lead to best remission.
For additional information (not part of the required reading), please see the following links:
Causal inference based on counterfactuals Read►
Causation, bias and confounding Read►
Causal analysis in epidemiology
- Counterfactual thinking deficit
- Counterfactual thinking in moral judgment
- Counterfactual reasoning and pretend play PubMed►
- The human disease network Read►
Causal reasoning in humans Read►
Farhan's lecture on backdoors Video►
This page is part of the course on Comparative Effectiveness by Farrokh Alemi, PhD Home►