{idmodelr}

# <code>{idmodelr}</code>
## An R package containing a library of infectious disease models and utilities
### Sam Abbott (👨‍🔬; <span class="citation">@seabbs</span>)
### University of Bristol
### 2019/11/16

---

# [`{idmodelr}`](http://bit.ly/idmodelr)<sup>1, 2</sup>

.footnote[
[1] Package Docs: [bit.ly/idmodelr](https://bit.ly/idmodelr), 
[2] JOSS draft paper: [bit.ly/idmodelr-paper-draft](https://bit.ly/idmodelr-paper-draft)
]

---
class: center, middle, inverse

# Slides: [bit.ly/idmodelr-satrday-cardiff](https://bit.ly/idmodelr-satrday-cardiff)

# `{idmodelr}`: [bit.ly/idmodelr](https://bit.ly/idmodelr)

# Dev Environment: [bit.ly/idmodelr-binder](https://bit.ly/idmodelr-binder)

---
class: inverse

# Outline

## 1. Who am I?

## 2. Why?

## 3. What?

## 4. How?

## 5. Future development

## 6. Wrap up

---
class: center, middle, inverse

# Who am I?

---

# Who am I?

* **Postdoctoral researcher:** Infectious disease dynamics.<sup>1</sup>

* **PhD:** Modelling tuberculosis and the optimal usage of the BCG vaccination.<sup>2</sup>

* **Previously:** Mathematics, theoretical physics, applied mathematics, and data science in finance.

* **Sideline:** Open source software, reproducibilty + hounding people on the internet about reproducibility.

* **Currently:** Job hunting.

[1] Thesis (in bookdown): [bit.ly/seabbs-thesis](http://bit.ly/seabbs-thesis)
]

---
class: center, middle, inverse

# Why?

---

# Why - domain area motivation

* Infectious disease models increasingly used by public health researchers but require a different knowledge base to other tools.

* Many resources exist for teaching theory but the R infectious disease modelling landscape is fragmented.

* Some modelling tasks are repetitive - these can easily be automated.

* Most tools focus on complex research usage. Examples:  [`pomp`](https://kingaa.github.io/pomp/), [`libbi`](http://libbi.org) and [`EpiModel`](http://www.epimodel.org).

* Several tools - notably [`{modelbuilder}`](https://github.com/ahgroup/modelbuilder) and [`{DSAIDE}`](https://ahgroup.github.io/DSAIDE/) - focus on teaching modelling dynamics.

* Most comparable to [`epirecipes`](http://epirecip.es/epicookbook/chapters/simple). 
---

# Why a 📦?

* A consistent framework on which to build.

* Tools to check the robustness of code (unit tests).

* Makes accessing models much easier and more consistent.

* Provides assurance that all models can be run in a single environment.

* Allows for tools to be developed alongside models to decrease the effort involved in modelling.

* Documentation on modelling can be provided alongside working, repurposable code examples.

---
class: center, middle, inverse

# What?

---

#Overview

* Library of infectious disease models (currently deterministic compartmental only - see future work).

* Easy to use (and robust) plotting and model summaries.

* Model and parameter look-up tables.

* Tools for scenario analysis.

* Still in active development with more models, tools, and docs in the pipeline.

---
class: center, middle, inverse

# Quick start

---

#0. Install

## Install the CRAN version

```r
install.packages("idmodelr")
```

## Alternatively install the development version

```r
# install.packages("devtools")
devtools::install_github("seabbs/idmodelr")
```

---

# 1. Find a model

```r
library(dplyr)
library(idmodelr)

idmodelr::model_details %>% 
  dplyr::filter(model_family %in% "SIR")
```

```
## # A tibble: 4 x 15
##   model model_family time  type  recovered exposed treated susceptible
##   <chr> <chr>        <chr> <chr> <chr>     <chr>   <chr>   <chr>      
## 1 SIR_… SIR          cont… dete… no        no      no      no         
## 2 SIR_… SIR          cont… dete… no        no      no      no         
## 3 SIR_… SIR          cont… dete… no        no      no      no         
## 4 SIR_… SIR          cont… dete… no        no      no      no         
## # … with 7 more variables: risk_stratified <chr>, non_exponential <chr>,
## #   simple_demographics <chr>, vaccination <chr>, disease_example <chr>,
## #   language <chr>, parameters <list>
```

.footnote[
Alternatively explore online: [samabbott.co.uk/idmodelr/articles/model_details.html](https://www.samabbott.co.uk/idmodelr/articles/model_details.html)
]

---

# 2. Check the model implementation

Look at the model + the help file (`?SIR_demographics_ode`).

```r
idmodelr::SIR_demographics_ode
```

```
## function (t, x, params) 
## {
##     S <- x[1]
##     I <- x[2]
##     R <- x[3]
##     with(as.list(params), {
##         N = S + I + R
##         dS = -beta * S * I/N - mu * S + mu * N
##         dI = beta * S * I/N - tau * I - mu * I
##         dR = tau * I - mu * R
##         derivatives <- c(dS, dI, dR)
##         list(derivatives)
##     })
## }
## <bytecode: 0x562b252c1f78>
## <environment: namespace:idmodelr>
```

---

# 3. Check the parameters

```r
parameters <- idmodelr::required_parameters("SIR_demographics_ode")

parameters
```

```
## # A tibble: 3 x 6
##   parameter parameter_family description type  risk_stratified
##   <chr>     <chr>            <chr>       <chr> <chr>          
## 1 beta      transmission     Transmissi… rate  no             
## 2 tau       recovery         Recovery r… rate  no             
## 3 mu        demographics     The natura… rate  no             
## # … with 1 more variable: non_exponential <chr>
```

.footnote[
Alternatively explore online: [samabbott.co.uk/idmodelr/articles/parameter_details.html](https://www.samabbott.co.uk/idmodelr/articles/parameter_details.html)
]

---

# 4. Parameterise the model

```r
parameters <- data.frame(
  beta = 3, ##Transmission rate = contact rate * transmission probablity
  tau = 0.5, ## Rate recovcery = 1 / duration of infection
  mu = 1/81 ## Natural birth/death rate = 1 / average lifespan
)
```

# 5. Initialise the model

```r
inits <- data.frame(
  S = 999,
  I = 1,
  R = 0
  )
```

---

# 6. Simulate the model

```r
## Model timespan
times <- seq(0, 50, 0.1)

## Simulate model as an ODE
traj <- idmodelr::simulate_model(model = SIR_demographics_ode,
                                 sim_fn = idmodelr::solve_ode, ##as solving an ode
                                 inits = inits,
                                 params = parameters,
                                 times = times)

head(traj, n = 5)
```

```
## # A tibble: 5 x 4
##    time     S     I      R
##   <dbl> <dbl> <dbl>  <dbl>
## 1   0    999   1    0     
## 2   0.1  999.  1.28 0.0567
## 3   0.2  998.  1.64 0.129 
## 4   0.3  998.  2.11 0.222 
## 5   0.4  997.  2.70 0.342
```

---

# 7. Summarise (WIP)

```r
idmodelr::summarise_model(traj) %>% 
  knitr::kable(format = "html")
```

<table>
 <thead>
  <tr>
   <th style="text-align:right;"> Final size: S </th>
   <th style="text-align:right;"> Final size: I </th>
   <th style="text-align:right;"> Final size: R </th>
   <th style="text-align:right;"> Epidemic peak time </th>
   <th style="text-align:right;"> Epidemic peak </th>
   <th style="text-align:right;"> Epidemic duration </th>
  </tr>
 </thead>
<tbody>
  <tr>
   <td style="text-align:right;"> 136 </td>
   <td style="text-align:right;"> 31 </td>
   <td style="text-align:right;"> 833 </td>
   <td style="text-align:right;"> 3.5 </td>
   <td style="text-align:right;"> 533 </td>
   <td style="text-align:right;"> Inf </td>
  </tr>
</tbody>
</table>

---

# 8. Plot (WIP)

```r
idmodelr::plot_model(traj, facet = FALSE)
```

---

# 9. Vary parameters and simulate

```r
parameters_up <- parameters
parameters_up[["mu"]] <- 1 / 20

traj_up <- idmodelr::simulate_model(
  model = idmodelr::SIR_demographics_ode,
  sim_fn = idmodelr::solve_ode,
  inits, 
  parameters_up,
  times)

head(traj_up, n = 5)
```

```
## # A tibble: 5 x 4
##    time     S     I      R
##   <dbl> <dbl> <dbl>  <dbl>
## 1   0    999   1    0     
## 2   0.1  999.  1.28 0.0565
## 3   0.2  998.  1.63 0.128 
## 4   0.3  998.  2.08 0.220 
## 5   0.4  997.  2.66 0.336
```

---

# 10. Compare models

```r
idmodelr::plot_model(traj, traj_up, facet = TRUE)
```

---
class: center, middle, inverse

# More here: [bit.ly/idmodelr](http://bit.ly/idmodelr)

---

# Not feeling like coding?

---

# `{shiny}` dashboard

```r
#install.packages("shiny")
#warning (eeek) this will install packages without a prompt on your system!
shiny::runGitHub("exploreidmodels", "seabbs")
```

---
class: center, middle, inverse

# How?

---

# Overview

* Simple code using the `{tidyverse}`,`{ggplot2}` + other 📦s.

* Package building tools including: `{devtools}`, `{usethis}`, `{testthat}`, `{vdiffr}`, and `{pgknet}`.

* R packages<sup>1</sup> by Hadley Wickham.

* Other 🛠s: git, Travis CI, Appveyor, codecov.io etc.

* Copying `{dplyr}`<sup>2</sup> religiously ......

* Aiming for peer review by the Journal of Open Source Software<sup>3</sup> from the get go.

.footnote[
[1] R packages 📖: http://r-pkgs.had.co.nz, [2] `{dplyr}`: https://github.com/tidyverse/dplyr,
[3] JOSS : https://joss.theoj.org/about
]

---

# More detail

---
class: center, middle, inverse

# Future development

---

# Planned features

* Additional models including: non-exponentially distributed, disease vector models, age stratified models, and many more.

* Stochastic implementations of models (plus more tooling to support).

* Models implemeneted using `Rcpp`, `LibBi` and `Stan`.

* S3 plot and summary methods.

* Model diagrams.

* Extending documentation and case studies.

* Streamlining and improving package tools (in particular scenario analysis tooling).

* Improving links with model fitting packages and post-processing packages (i.e cost effectiveness).

---

# Please join in<sup>1</sup>!

.footnote[
[1] Issues: [github.com/seabbs/idmodelr/issues](https://github.com/seabbs/idmodelr/issues)
]
---
class: center, middle, inverse

# Why join in?

---
class: center, middle, inverse

## [`{idmodelr}`](http://bit.ly/idmodelr) = maths + data + #rstats + a global (preventable) issue

## -> [`{idmodelr}`](http://bit.ly/idmodelr) = the best thing since sliced bread<sup>*</sup>

---

# Why join in?

* `{idmodelr}`  contains perhaps the simplest possible code for understanding infectious disease models and so may be a good place to start learning more.

* For subject area experts contributing may help link up the package with your work.

* Contributing to packages is a great way to learn about package development. No domain knowledge is required.

* Heading towards a JOSS publication in 2020 - authorship up for grabs (a bribe for academics).

---

# Wrap up

---

# Summary

**Why**

* Infectious disease models are widely used both by domain area experts and non-experts.

* For non-experts, existing tooling and teaching material can be complex, difficult to get started with or fragmented.

* For experts, routine tasks can be easily automated.

**What**

* A library of infectious disease models.

* Tools to use them.

* Documentation explaining what is going on.

* A `{shiny}` app to get users started quickly with no code.

* A focus on keeping model code simple, easy to find, and easy to modify.

---
class: center, middle, inverse

# Now for some advertising

---

## Introduction to infectious disease modelling course

.footnote[
[1] Website: [bristolmathmodellers.github.io/biddmodellingcourse/](https://bristolmathmodellers.github.io/biddmodellingcourse/),

[2] Sign up: [http://bit.ly/bristol-idm-course](http://bit.ly/bristol-idm-course)
]

---

# `{getTBinR}`<sup>1</sup>

---

# CV using `#rstats`<sup>1</sup>...

.footnote[
[1] CV: [bit.ly/seabbs-cv](http://bit.ly/seabbs-cv), Code: [github.com/seabbs/cv](https://github.com/seabbs/cv), Talk: [bit.ly/seabbs-CVinR](http://bit.ly/seabbs-CVinR)
]

---
class: inverse, center, middle

# Thanks for listening!

## Slides: [bit.ly/idmodelr-satrday-cardiff](http://bit.ly/idmodelr-satrday-cardiff)

## Code: [bit.ly/idmodelr-satrday-cardiff-code](http://bit.ly/idmodelr-satrday-cardiff-code)

## 📦 : [bit.ly/idmodelr](http://bit.ly/idmodelr)

## 👨‍🔬 : [bit.ly/seabbs](http://bit.ly/seabbs)

## Tweet at me: [@seabbs](https://twitter.com/seabbs)