report.Rmd

---
header-includes:
   - \usepackage{float}
output:
  pdf_document: default
  html_document: default
  citation_package: natbib
  number_sections: true
  extra_dependencies: ["caption", "subcaption", "graphicx"]
  toc: true
---

```{r, echo=FALSE, message=FALSE}
message("~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ running cor_threshold report ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~")
```


```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
setwd("/Users/larsvanderlaan/correlates_reporting_forks/correlates_reporting")
library(here)
library(dplyr)
print(here::here())
source(here::here( "_common.R"))
 
times <- intersect(c("Day57", "Day29"), times)
keys_short <- times
if(study_name=="ENSEMBLE" | study_name=="MockENSEMBLE") {
  keys_short <- c(keys_short, "Day29start1")
}
keys <- c()
markers <- markers
markers <- unlist(sapply(times, function(v) grep(v, markers, value = T))) # Remove the baseline markers
for(assay in assays) {
  keys <- c(keys, paste0(keys_short, assay))
}
 
key_to_markers <- markers
names(key_to_markers) <- markers


marker_to_time <- sapply(markers, function(v) {
  times[stringr::str_detect(v, times)]
})
marker_to_assay <- sapply(markers, function(v) {
  unname(assays[stringr::str_detect(v, assays)])
})


key_to_time <- sapply(keys, function(v) {
  times[stringr::str_detect(v, times)]
})
key_to_assay <- sapply(keys, function(v) {
  unname(assays[stringr::str_detect(v, assays)])
})
key_to_markers <- paste0(key_to_time, key_to_assay )
names(key_to_markers) <- keys
key_to_markers
key_to_short <- sapply(keys, function(v) {
  tmp <- setdiff(keys_short, "Day29start1")
  if(stringr::str_detect(v, "Day29start1")) {
    return("Day29start1")
  } else {
    return(unname(tmp[stringr::str_detect(v, tmp)]))
  }
   
})
```

# Univariate CoR: Nonparametric Threshold Modeling (>=s)

An extension of the unadjusted nonparametric threshold-searching approach
developed in @Donovan, the covariate-adjusted TMLE-based approach developed by
van der Laan, Zhang, Gilbert (submitted) is used to estimate the so-called
threshold-response function $E_X[ E[Y \mid S \geq s, X, A = 1] \mid A=1]$ for a
range of thresholds $s$.Here, $X$ is a set of baseline characteristics, $A=1$
represents the vaccine group, $S$ is the biomarker/immune-response/correlate of
interest, and $Y$ is the indicator of COVID disease before some time point
$t_f$. This parameter can be viewed as a causal/covariate-adjusted version of
the parameter $P(Y=1 \mid S \geq s, A=1)$. Intuitively, the threshold-response
at a given threshold is the expected probability of obtaining COVID disease if
one experiences a marker/immune-response value above that threshold. The
threshold-response function is estimated for each of the four Day 57 antibody
markers, in each case adjusting for the baseline covariates: baseline risk
score, high risk indicator, and underrepresented minority status. A restrictive but flexible specification of the Highly Adaptive Lasso estimator is used for the covariate adjustment. A
number of plots and tables are reported:
\begin{enumerate}
\item A plot and table with risk estimates and point-wise 95\% confidence intervals
\item A plot and table with risk estimates and simultaneous 95\% confidence bands
\item Monotone-corrected versions of 1 and 2.
\end{enumerate}

A reverse cumulative distribution function curve estimated by the IPW NPMLE of
the marker values is superimposed on the threshold-response plots and a dashed
red line is added to mark the threshold value after which no more events are
observed.

The blue dots on the plots represent the risk predictions at marker values where there was an observed COVID-19 case.


```{r, echo = F, include = F}
short_keys <- unique(key_to_short)

```



```{r, echo = F, include = F}
days <- c()
if(has57) {
  days <- c(days, 59)
}
if(has29) {
  days <- c(days, 29)
}

ci <- "pointwise"
above = TRUE
monotone <- FALSE

```


```{r, results="asis", echo = F }
 
for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    cat(key)
    #res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    res <- ""
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}


monotone <- TRUE
for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    cat(key)
    #res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}



```



```{r, echo = F , results = "asis", eval = F}
ci <- "simultaneous"

for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}


monotone <- TRUE
for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}

```









# Univariate CoR: Nonparametric Threshold Modeling (<=s)

The same methodology as the previous section is apply to estimate the "below" threshold-response function $E_WE[Y=1|A=1,X,S \leq s].$



```{r, echo = F, include = F, eval = F}
 
ci <- "pointwise"
above = FALSE
monotone <- FALSE

```


```{r, results="asis", echo = F, eval = F}
setwd("/Users/larsvanderlaan/correlates_reporting_forks/correlates_reporting")
for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}


monotone <- TRUE
for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}



```



```{r, echo = F , results = "asis", eval = F}
ci <- "simultaneous"

for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}


monotone <- TRUE
for(key in short_keys){
  countAfter1 <- FALSE
  if(length(stringr::str_detect(key,"start"))>0) {
    countAfter1 <- TRUE
  }
  day <- gsub("Day", "", key)
  day <- gsub("start1", "", key)
  title <- paste0("Plots and Tables with estimates and ", ci, " confidence intervals"," for Day ", day)
  if(monotone) {
    title <- paste0(title, " (monotone-corrected)")
  }
  if(countAfter1) {
    title <- paste0(title, " (counting events after 1 day post dose)")
  }
  cat("##", title, "\n")
  for(assay in assays) {
    
   key <- paste0(short_keys, assay)
    marker <- key_to_markers[key]
    res = knitr::knit_child(c(here::here('cor_threshold', 'sub_report.Rmd')), quiet = TRUE)
    cat(res, sep = '\n')
    cat("\\clearpage", sep = '\n')
  }
}

```