battenr

# Title: Propensity Score Plots Using IP Weighting ----

# Description: This code shows how different patients will receive different weights
# depending upon the causal estimand when using IP weighting. This code focuses on 
# three estimands: ATE, ATT and ATU. 

# Note: thanks to Andrew Heiss' blog for the inspiration for these plots and making his code available!

# Setup ----

battenr

# Title: M-Estimation After IPTW ----

# Description: Demonstrating M-estimation after IPTW. 
# Showing how it can be helpful, and comparing to if we didn't use anything 
# (in this case using the SEs from GLM)

# Note: code in this script was largely inspiried by Ross et al. (2024). 
# The code supplied in that paper was used to create this code 
# (the authors used a binary outcome, while this code uses a continuous outcome)

battenr

# Title: Understanding Diagnostic Plots Through Simulation ----

# This code uses an example to demonstrate how we can use simulation to 
# better understand diagnostic plots. For this specific example, we will use the 
# posterior predictive check 

# Note: using default priors, etc. for this example

# Setup: ----

battenr

# Title: Table 2 Fallacy ----

# Description: Demonstrating the table 2 fallacy using a simple example.  
# This is based on the DAG that is shown in a corresponding LinkedIn post. 

# Setup ----

#... Libraries ----

library(tidyverse) # ol faithful 

battenr

# Title: Time-Varying Confounding (MSM vs IPTW)

# Description: Demonstrating how marginal structural models provide 
# less biased estimates compared to using IPTW at baseline only, when 
# there is time-varying confounding. 

# Setup ----

#... Libraries ----

battenr

# Title: Matching vs Weighting 

# Description: Demonstrating how matching and weighting can estimate 
# different things. IPTW can also estimate ATT. The goal is to demonstrate how they 
# may give different answers

# Setup ----

#... Library ----

battenr

# Title: Doubly Robust Estimation

# Description: Demonstrating how doubly robust estimation can be 
# a useful tool, but has limitations. Specifically by looking at three different 
# scenarios and the resulting bias. 

# This script has three examples: 
# 1. Both models correctly specified
# 2. One of the models correctly specified
# 3. Neither model correctly specified. 

battenr

# Title: Ordered Beta Regression 

# Description: This code is to demonstrate using ordered beta regression. 
# This method can be particularly helpful for continuous data with upper/lower bounds. 

# For more information recommend reading Kubinec (2022)

# Setup ----

#... Packages ----

battenr

# Title: Plotting Different Priors 

# Description: Choosing an appropriate prior is a key part of any Bayesian analysis. 
# This code shows different priors to try and highlight which one(s) might 
# be reasonable. 

# Setup ----

#... Packages ----

battenr

# Title: Change from Baseline 

# Description: Analyzing change from baseline requires some additional assumptions. 
# There are seven laid out in the link below from Frank Harrell
# Highly recommend reading it

# https://www.fharrell.com/post/errmed/#change-from-baseline

# This post aims to demonstrate one of those issues: 
# To use change from baseline the pre-post measurements need to be linearly related.