battenr

# Title: Stabilizing and Truncating Weights 

# Description: When using weights, for example IPTW, it is often helpful to 
# stabilize and/or truncate them. This code is meant to demonstrate how we 
# can do that, and why. 

# Setup ----

#... Packages ----

battenr

# Title: Checking Propensity Scores and Covariates

# Description: This script demonstrates that sometimes there can appear to be overlap when checking only the propensity score, 
# however upon checking the covariates overlap, there is a lack of overlap. 

# Setup ----

#... Libraries ----

library(tidyverse) # ol' faithful

battenr

# Title: Strength of Confounders When Choosing a Data Source

# Description: When using real-world data, an important aspect is choosing the 
# dataset to use. Not all variables are always available in each data source. 
# For example, eletronic health records vs claims data vs registry data each 
# have strengths and benefits (one of which, is the variables available)

# Setup ----

#... Packages ----

battenr

# Title: Adjusting for Predictor Variable

# Description: Directed acyclic graphs (DAGs) are fantastic for variable selection. 
# Sometimes there are variable that it's not clear if we should include or not. 
# This code demonstrates one example: a variable that causes one of the confounders
# and the outcome. 

# Both bias and relative increase in precision are explored for this situation. 
# Inverse probability weighting, estimating the average treatment effect, is 
# used. When adjusting for the predictor, it is only included in the outcome model

battenr

# Title: What to Do If Lack of Overlap? 

# Description: This code creates a figure where there is a lack of overlap 
# due to leaving out a confounder. 

# Setup ----

#... Libraries ----

library(tidyverse) # ol' faithful

battenr

# Title: Collider and confounder - adjust or not adjust? (Butterfly bias)

# Description: Understanding if we should adjust for a variable that is 
# both a collider and a confounder. For this example, we will use a DAG that has 
# butterfly bias (a specific type of M-bias)

# The idea for this code was to keep the situation simple to start. It can be built upon 
# rather easily. A few things to try altering would be the strength of the causal relationships, 
# continuous vs binary vs TTE variables, more/less variables, etc. 

battenr

# Title: Adjusting for Mediator ----

# Description: This code demonstrates how adjusting for a mediator doesn't 
# estimate the total causal effect. 

# Setup ----

library(tidyverse) # ol' faithful
library(ggdag) # for drawing DAGs
library(broom) # tidying model outputs 

battenr

# Title: Checking Balance After IPTW: SMDs & Plots

# Description: This code demonstrates how standardized mean differences (SMDs) 
# can be a useful tool to check for balance but may not show the entire picture. 

# Setup ----

#... Packages ----

library(tidyverse) # ol faithful

battenr

# Title: Estimating Sample Size Required Using Precision -----

# Description: Sample sizes are primarly based upon an assumed power. For some studies, 
# such as observational studies with large samples, there isn't an issue with power. However, 
# there may be a need to understand the required sample size (i.e., how much data to request from vendor, etc)

# In these situations, precision can be used instead! 

# Setup ----

battenr

# Doubly Robust Estimation ----

# Description: Doubly robust estimation can be helpful for estimating 
# causal effects. This code is demonstrating how there is reduced bias
# compared to IPTW only or using the outcome model only

# Setup ----

#... Packages ----