sandbox.R

In asshah4/octomod: A Causality-Informed Modeling Approach

# Sandbox ----
library(tidyverse)
library(survival)
library(gt)

# gt tables ----
clinical <- readRDS("../mims/_targets/objects/clinical")
models <- readRDS("../mims/_targets/objects/cox_models")
outcomes <- readRDS("../mims/_targets/objects/outcomes")
cox <- readRDS("../mims/_targets/objects/cox_compare")

data = extract_results(cox, how = "tidy", flat = TRUE, exponentiate = TRUE, conf.level = 0.95, conf.int = TRUE)
terms = term ~ list(
	lf_stress = "HRV",
	lf_rest = "HRV",
	hf_stress = "HRV",
	hf_rest = "HRV",
	bpm_rest = "BPM",
	bpm_stress = "BPM",
	rdr_msi_bl1 = "MSIMI",
	lntroponin_rest = "Troponin"
)
models = name ~ list(
	death_base = "Unadjusted",
	death_msimi = "MSIMI",
	death_trop = "Troponin"
)
statistic = p.value ~ 0.05
values = c("estimate", "conf.low", "conf.high")
pattern = "{1} ({2}, {3})"
style = fill ~ list(color = "lightgreen")
decimals = 2
by = exposures ~ list(
	lf_stress = "Stress LF",
	lf_rest = "Rest LF",
	hf_stress = "Stress HF",
	hf_rest = "Rest HF",
	bpm_rest = "Rest Pulse",
	bpm_stress = "Stress Pulse"
)
missing_text = "."

tbl_compare(
	data = data,
	terms = terms,
	by = by,
	models = models,
	statistic = statistic,
	values = values,
	pattern = pattern,
	style = style,
	decimals = decimals,
	missing_text = missing_text
)

# The map tibble should only be used for a single hypothesis at a time.
# Grouping/order:
	# 1. outcome
	# 2. exposure
	# 3. covariate
data <- subset(object, name == "hseq")

data %>%
	select(outcomes, exposures, number, term, estimate) %>%
	pivot_wider(
		names_from = outcomes,
		values_from = estimate,
	) %>%
	gt(rowname_col = "term", groupname_col = "exposures")