README.Rmd

---
output: github_document
---

<!-- README.md is generated from README.Rmd. Please edit that file -->

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
options(width = 200)
```

# My modified OncoBayes2

<!-- badges: start -->
<!-- badges: end -->

This is my modified version of the [original OncoBayes2](https://CRAN.R-project.org/package=OncoBayes2) package.

The most significant difference is the support of providing correlation parameter when sampling the component
via the newly added parameter `prior_EX_corr_mu_comp` in `blrm_exnex()`.

In the original package, $\theta = (log \alpha, log \beta) \sim N(\mu, \Sigma)$ where the __covariance__ matrix $\Sigma$ is a diagonal matrix whose diagonals are provided via `prior_EX_mu_sd_comp`(Note, this parameter provides the __standard variance__).

Now, the off-diagonal part of $\Sigma$ can also be specified by supplying the __correlation__ via `prior_EX_corr_mu_comp`.

## NOTE

1. This is just a __proof of concept__ during my replication of Table 14-7 in this [trial protocol](https://clinicaltrials.gov/ProvidedDocs/64/NCT02108964/Prot_001.pdf). The original OncoBayes2 misses the functionality
to accept user defined correlation, which is also reported on the [internet](https://stats.stackexchange.com/questions/617633/correlation-for-priors-in-blrm-with-ewoc-in-r-package-oncobayes2/626132#626132)

1. Currently this package is forked and modified from [CRAN's github version of OncoBayes2](https://github.com/cran/OncoBayes2). After installation it will __overwrite__ your existing OncoBayes2. 

1. This is currently WIP. Many aspect, like the `prior_summary` method should be updated accordingly to reflect the correlation information. Also many document template is missing hence the manual cannot be auto-updated properly.

## Installation

You can install the development version of my modified OncoBayes2 like so:

``` r
remotes::install_github("fenguoerbian/MyOncoBayes2")
```

## Example

This is a basic example which reproduce the results in Table 14-7 in this [trial protocol](https://clinicaltrials.gov/ProvidedDocs/64/NCT02108964/Prot_001.pdf).

```{r example}
library(OncoBayes2)
```

### Setup MCMC options

```{r}
set.seed(123)
.user_mc_options <- options(OncoBayes2.MC.warmup=500, OncoBayes2.MC.iter=20000, OncoBayes2.MC.chains=10,
                            OncoBayes2.MC.save_warmup=FALSE, mc.cores = 10)

```

### Setup basic trial infomation

```{r}
SA_trial_setup <- blrm_trial(
    data = tibble::tibble(group_id = as.factor("All"),
                  drug_A = 50,
                  num_patients = NA,
                  num_toxicities = NA),
    # dose-toxicity data available at design stage of trial.
    
    dose_info = tibble::tibble(group_id = as.factor("All"), 
                       drug_A = c(50,75,150,300,450,600,800,1000), 
                       dose_id = c(1, 2, 3, 4, 5, 6, 7, 8), stratum_id = "all"), 
    # specification of the dose levels as planned for the ongoing trial arms.
    
    drug_info = tibble::tibble(drug_name = "drug_A", 
                       dose_ref = 300, 
                       dose_unit = "ug/kg"
                       ,
                       reference_p_dlt = exp(-3.068) / (1 + exp(-3.068))
    )
    # specification of drugs used in trial arms
    
    ,simplified_prior = FALSE
    ,EXNEX_comp=TRUE
    ,EXNEX_inter=FALSE
    ,interval_prob = c(0,0.16,0.33,1)
    ,interval_max_mass = c(prob_underdose = 1 # The prob of under-dose is allowed maximum to 1.
                           , prob_target = 1 # The prob of target-dose is allowed maximum to 1.
                           , prob_overdose = 0.25) # The prob of over-dose is allowed maximum to 0.28.
)


dims <- summary(SA_trial_setup,"dimensionality")
num_comp <- dims$num_components
```

### Update the prior

```{r, message=FALSE}
dims <- summary(SA_trial_setup,"dimensionality")
num_comp <- dims$num_components

SA_trial_start <- update(
    SA_trial_setup,
    # component1 MAP prior
    # check the usage in single agent example documentation.
    prior_EX_mu_mean_comp = matrix(
        c(-3.068, # mean of intercept 
          0.564), # mean of log-slope 
        nrow = num_comp,
        ncol = 2
    ),
    prior_EX_mu_sd_comp = matrix(
        c(2.706, # sd of intercept
          0.728), # sd of log-slope
        nrow = num_comp,
        ncol = 2
    ),
    prior_EX_corr_mu_comp = -0.817,    # HERE! supply the CORRELATION!
    prior_EX_tau_mean_comp = matrix(
        c(0, 0),
        nrow = num_comp,
        ncol = 2
    ),
    prior_EX_tau_sd_comp = matrix(
        c(1, 1),
        nrow = num_comp,
        ncol = 2
    ),
    prior_EX_prob_comp = matrix(1, nrow = 1, ncol = 1),
    prior_tau_dist = 0, # single-agent, without historical data, set this to 0.
                        # see single-agent example in this package for more details
    prior_PD = FALSE
)
```

__NOTE:__ `prior_summary` needs to be updated to reflect user supplied correlation

```{r}
prior_summary(SA_trial_start)
```

### Check the results against Table 14-7

```{r}
summary(SA_trial_start, "dose_prediction")
```

```{r}
summary(SA_trial_start, "ewoc_check")
```