model-summary.Rmd

---
title: "Model Summary"
output: html_document
---

```{r libraries, include = FALSE}
library(bbr)
library(tidyverse)

source("functions-model.R")

knitr::opts_chunk$set(
  out.width = 500, 
  out.height = 750
)
```

```{r directories, include=F}
### Directories ----------------------------
scriptDir <- getwd()
projectDir <- file.path("..")
figDir <- file.path(projectDir, "deliv", "figure")
tabDir <- file.path(projectDir, "deliv", "table")
```

```{r echo = FALSE, results = "asis"}
cat("## Run log")
```

```{r }
# create the raw run log
MODEL_DIR <- "../model/pk"

log_df <- run_log(MODEL_DIR) %>% 
      add_config() %>%
      add_summary()
```

```{r echo = FALSE, results = "asis"}
# These two are helpers for doing some
# automatic QC on all the models in log_df

# check if model or data files have changed (from bbr)
check_up_to_date(log_df)

# check if any rogue tags that aren't in tags.yaml (from functions-model.R)
tags_ok <- audit_tags(log_df, yaml::read_yaml(here::here("script", "tags.yaml")))
if (all(tags_ok)) cat("All tags are valid.\n")
```

```{r}
# format log_df for display
log_df %>% 
  collapse_to_string(based_on, tags, notes) %>%
  select(run, based_on, description, tags, ofv, param_count, notes) %>%
  knitr::kable()
```

### Tags diff

_Optionally view the tags that were added or removed for each model._

```{r}
# format log_df for display
log_df %>% 
  add_tags_diff() %>%
  format_tags_diff() %>%
  select(run, based_on, description, tags, tags_diff) %>%
  knitr::kable()
```

# Modeling notes

## Model 100-101

Compare Model 100 (one-cmpt) and 101 (two-cmpt)

```{r obj_100_101}
# this is a helper from functions-model.R which filters
# the summary log tibble to only the specified columns
compare_ofv(100:101, log_df)
```

\newpage
Compare CWRES between model 100 & 101
```{r 100_101_cwres_time_tad}
knitr::include_graphics(file.path(figDir, "100/100-cwres-pred-time-tad.pdf"))
knitr::include_graphics(file.path(figDir, "101/101-cwres-pred-time-tad.pdf"))
```

Second compartment reduces the objective function and residual issues.

\newpage

# Model 101-102

Look at ETA's vs. covariates to decide how to begin incorporating covariates.

```{r}
mod101 <- read_model(file.path(MODEL_DIR, 101))
print(mod101$notes)
```

```{r, out.extra='page=2'}
knitr::include_graphics(file.path(figDir, "101/101-eta-all-cont-cov.pdf"))
```

```{r, out.extra='page=2'}
knitr::include_graphics(file.path(figDir, "102/102-eta-all-cont-cov.pdf"))
```

## Models 102-104
Proportional, additive or combined RUV

```{r}
compare_ofv(102:104, log_df, notes)
```

\newpage
## Models 105-106
105 adds pre-specified covariates. For 106, client was interested in adding Albumin to CL.
  
```{r}
# filter the run log to only models with this covariate tag
TAGS <- yaml::read_yaml("tags.yaml")

run_log(MODEL_DIR, .include = TAGS$cov_cl_egfr) %>% 
  add_summary() %>%
  select(run, ofv, param_count, condition_number) %>%
  knitr::kable()
```

# Comparing final model to base model

```{r}
log_df %>% 
  filter(star) %>%
  select(run, description, ofv, param_count) %>%
  knitr::kable()
```

```{r}
mod102 <- read_model(file.path(MODEL_DIR, 102))
mod106 <- read_model(file.path(MODEL_DIR, 106))

tags_diff(mod102, mod106)
```

```{r, results = "asis"}
model_diff(mod102, mod106)
```


# Final model - 106

```{r}
mod106 %>%
  model_summary() %>%
  param_estimates() %>%
  mutate(across(where(is.numeric), pmtables::sig)) %>%
  knitr::kable()
```