How do I add subscripts to labels in ggplot? - r

Question

I'm doing an analysis on air pollutants using Bayesian Kernel Machine Regression, using the bkmr package in R.
https://jenfb.github.io/bkmr/overview.html
The link is to Jennifer Bobb's instructions on how to use this package. I don't think it is relevant to the issue though. What I want to do is have PM2.5, O3, and NO2 show up in my charts with the 2.5, 3, and 2 as subscripts. I'm trying to use this function and getting no luck:
colnames(dat) <- c("LTE4", "$O[3]", "$PM[2.5]", "$NO[2]", "Diethyl", "Dimethyl", "age", "tmpf", "relh", "sex", "agany", "agself", "asthma")
When I do this what happens I just see these labels show up in the plots with with the $ and [] instead of subscripted numbers. Any ideas?
This is the full code I am using:
### January BKMR Analysis ###
## Hierarchical Variable Selection ##
## Updated June 6, 2022 ##
# Reading in necessary packages
library(tidyverse)
library(bkmr)
trio_semipro <- readRDS("C:/Users/Matt/OneDrive/Documents/Fresno Thesis/Thesis Code/trio_semipro.rds")
trio_semipro
dim(trio_semipro)
head(trio_semipro)
trio_semipro$log_lte4 <- log(trio_semipro$Final)
# Separating out dataframes for winter and summer to run separate models for each season
trio_semipro_w <- trio_semipro %>%
filter(visit_month == 1)
trio_semipro_s <- trio_semipro %>%
filter(visit_month == 2)
# Summer and Winter Dataframes
trio_semipro_w
trio_semipro_s
head(trio_semipro_w)
#view(trio_semipro_w)
dat = cbind(trio_semipro_w$log_lte4, trio_semipro_w$O3,
trio_semipro_w$PM25, trio_semipro_w$NO2, trio_semipro_w$diethyl, trio_semipro_w$dimethyl,
trio_semipro_w$age, trio_semipro_w$tmpf, trio_semipro_w$relh, trio_semipro_w$sex, trio_semipro_w$agriculture_anyone,
trio_semipro_w$agriculture_self, trio_semipro_w$asthma)
head(dat)
colnames(dat) = c("LTE4", "$O[3]", "$PM[2.5]", "$NO[2]", "Diethyl", "Dimethyl", "age", "tmpf", "relh", "sex", "agany", "agself", "asthma")
dat = as.data.frame(dat)
dat$sex
# recode the binary variable to be 0, 1 and NA
dat$agself = dat$agself-1
dat$agself[which(dat$agself==2)]=NA
dat$agself
# recode sex variable
dat$sex = dat$sex -1
# recode agany variable
dat$agany = dat$agany - 1
dat$agany[which(dat$agany==2)]=NA
#recode asthma variable
dat$asthma = dat$asthma - 1
dat$asthma[which(dat$asthma==2)]=NA
dat$asthma
dat$sex
dat$agany
# good
head(dat)
complete_dat = dat[-which(apply(dat, 1, anyNA)),]
dim(complete_dat)
# Fit BKMR
zscaled <- apply(complete_dat[,(2:6)], 2, scale)
yscaled <- scale(complete_dat$lte4)
xscaled <- cbind(scale(complete_dat[,7:9]), complete_dat[,10:13])
fit_bkmr = kmbayes(y=yscaled, Z= zscaled, X = xscaled,
iter = 20000, varsel = TRUE, groups=c(1,1,1,2,2), verbose=FALSE)
plot(fit_bkmr$sigsq.eps, type = "l")
TracePlot(fit = fit_bkmr, par = "beta", comp = 4)
TracePlot(fit = fit_bkmr, par = "sigsq.eps")
TracePlot(fit = fit_bkmr, par = "r", comp = 1)
# Estimating posterior inclusion probabilities
ExtractPIPs(fit_bkmr)
# Estimating h
y <- yscaled
Z <- zscaled
X <- xscaled
med_vals <- apply(Z, 2, median)
Znew <- matrix(med_vals, nrow = 1)
# Summarize model output
pred.resp.univar <- PredictorResponseUnivar(fit = fit_bkmr)
library(ggplot2) # Using ggplot to plot cross sections of h
ggplot(pred.resp.univar, aes(z, est, ymin = est - 1.96*se, ymax = est + 1.96*se)) +
geom_smooth(stat = "identity") +
geom_hline(yintercept = 0, lty = 5, col = "red2", alpha = 0.4) +
facet_wrap(~ variable, nrow = 1) +
ylab("h(z)")
# visualze the bivarate exposure-response function for two predictors, where
# all of the other predictors are fixed at a particular percentile.
pred.resp.bivar <- PredictorResponseBivar(fit = fit_bkmr, min.plot.dist = 1)
ggplot(pred.resp.bivar, aes(z1, z2, fill = est)) +
geom_raster() +
facet_grid(variable2 ~ variable1) +
scale_fill_gradientn(colours=c("#0000FFFF","#FFFFFFFF","#FF0000FF")) +
xlab("expos1") +
ylab("expos2") +
ggtitle("h(expos1, expos2)")