Causal Analysis using R Interrupted Time Series Analysis

Readme.txt

 This R snippets shows how to perform interrupted time series analysis on a synthetic data set using R.

InterrupedTimeSeries.R

#### Causal Analysis using R Interrupted Time Series

### Brief Overview of ITS
#  - Interrupted time series (ITS) is a quasi-experimental method used to evaluate the impact 
#    of an intervention by analyzing an outcome measured over time before and after the 
#    intervention occurs.
#  - The technique works by modeling changes in both the level and slope of the time series 
#    that follow the intervention, estimating both immediate and long-term effects.
# -  ITS is valuable when randomization is not feasible and can provide strong evidence 
#    for causal inference, but requires multiple data points on both sides of the intervention.
# -  Key assumptions include that the pre-intervention trend would have continued in the 
#    absence of the intervention, and that no major confounding events coincide with 
#    the intervention period.
# -  ITS is frequently used in public health, policy, and social sciences to assess population-level 
#    changes resulting from interventions or policy changes.
 
install.packages("its.analysis")

library(its.analysis)


# Generating Synthetic Data
year <- c(2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008,
          2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018)
depv <- c(8.22, 8.19, 8.23, 8.28, 8.32, 8.39, 8.02,
          7.92, 7.62, 7.23, 7.1, 7.11, 6.95, 7.36, 7.51, 7.78, 7.92, 7.81)
interruption <- c(0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0)
cov1 <- c(3.1, 3.3, 5.1, 5.2, 5.4, 4.5, 4.7, 4.9, 5.3,
          5.6, 5.8, 6.0, 4.8, 5.2, 4.5, 4.6, 5.1, 4.7)

# Create data frame
x <- as.data.frame(cbind(year, depv, interruption, cov1))
head(x)

# Run interrupted time series model
itsa.model(data=x, time="year", depvar="depv", interrupt_var="interruption",
           alpha=0.05, bootstrap=TRUE, Reps=250)

##Interpretations
##ANOVA Table   the depvar significantly associated with interrupt_var (intervention effects)
##              the lag_depvar highly significant (dependency on prior data points)
##Tukey. result pairwise group comparison controlling for familywise type I error.
##      diff shows direction (in this case: intervention caused a decine in depvar)
##itsa.result a string of a statement of conclusions
##Shapiro.test for normality of data; levenes.test for equal variance
##autcorr for autocorrelation of residuals   

# Run model with covariate and different alpha level
itsa.model(data=x, time="year", depvar="depv", interrupt_var="interruption",
           covariates="cov1", alpha=0.1, bootstrap=TRUE, Reps=250)

# Run model without plots and bootstrap.(In this case, linear regression is usually used)
# Bootstrap is preferred when the data distribution is not normal or sample size is limited
itsa.model(data=x, time="year", depvar="depv", interrupt_var="interruption",
           covariates="cov1", alpha=0.1, no.plots=TRUE, bootstrap=FALSE)

The link to the original YouTube video: https://www.youtube.com/watch?v=3-dN19Vhkiw&list=PL3OtMQSLGztmePsY5aXjUiYBrP1Sw8ml7