Refined version of adding Standard Error Bars to Kenny et al. original graphs

[yuexinz04]

Refined version of adding Standard Error Bars

1. Full GAM model

gam(error ~ t2(turnout, dem, log(dens)) + s(white) + s(hh), data = d)

For projecting the standard error bars of each point of the fitted error in Kenny et al. original graphs, instead of using the "geom_errorbar", use "geom_pointrange".

Add-on to ggplot()
{
geom_pointrange(aes(ymin=fitted(m)-1.96se, ymax=fitted(m)+1.96se), width=.1) +
}

• Plot [Fitted Error Vs Non-White Percent]
  -->"refine_addRange_race_error.pdf"
• Plot [Fitted Error Vs HHI]
  --> "refine_addRange_herfindahl_error.pdf"

3. Reduced GAM model

Plot the fitted error (response variable) against distinct predictor variable (Herfindahl Index or Non-White Percent).
Project the aforementioned standard error bars of each point of the fitted error to the graphs.

• gam (error ~ s(hh),data=d)
  -->Plot [Fitted Error Vs HHI]
  --> "gamHHI_herfindahl_error.pdf"

• gam(error ~ s(white), data = d)
  --> Plot [Fitted Error Vs Non-White Percent]
  --> "gamNonWhite_race_error.pdf"

[yuexinz04]

Complete

https://github.com/yuexinz04/2022-Summer-Statistical-Privacy-and-Differential-Privacy/tree/kennyetal_modification/Refined_addRange