Visualising experimental designs with the edibble and deggust R-packages

.info-box.w-50.bg-white[
These slides are viewed best by Chrome or Firefox and occasionally need to be refreshed if elements did not load properly. Please note this web version takes a while to load. See <a href="deggust.pdf">here for the PDF </a>. 
]

---

background-size: cover
class: title-slide
count: false

# .f1.monash-blue[Visualising experimental designs with the `edibble` and `deggust` R-packages]

.bottom_abs.width100[

Presenter: *Emi Tanaka*

Department of Econometrics and Business Statistics, &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Monash University, Melbourne, Australia

emi.tanaka@monash.edu

<a href="https://twitter.com/statsgen" style="color:black"> @statsgen</a>

11 Nov 2021 @ Applications of Statistical Procedures in Biological Data

]

---

.f1.tl[
 .menu-font[today's menu]]

.f2.pa3.tl[
.plate[1] Experimental design in *reality* .f3.grey[] [.f4[]](#reality)

]

.border-box.monash-bg-blue2.center.w-70[
 These slides made using R powered by HTML/CSS/JS can be found at <a class="f2" style="color:yellow!important" href="https://emitanaka.org/slides/stats4bio2021/deggust">emitanaka.org/slides/stats4bio2021/deggust</a>
]

---

# Experimental design in *reality*

---

# .f-headline[An experiment generally involves *more than one person*]

---

The **"domain expert"**
]
.w-33[
<img src="images/stickman/statistician1.svg" width="120px" class="">

The **"statistician"**

The **"analyst"**
]
.w-33[
 
<img src="images/stickman/technician2.svg" width="100px" class="img-hor">

The **"technician"**
]
]

.absolute.bottom-0.f5[
Stick person images by <a href="https://pixabay.com/users/openclipart-vectors-30363/?utm_source=link-attribution&amp;utm_medium=referral&amp;utm_campaign=image&amp;utm_content=151793">OpenClipart-Vectors</a> from <a href="https://pixabay.com/?utm_source=link-attribution&amp;utm_medium=referral&amp;utm_campaign=image&amp;utm_content=151793">Pixabay</a>
]

---

.f-subheadline.hidden[Meet the cast starring today]

The **"domain expert"**
]
.w-60.tl[

The .monash-blue2[**"domain expert"**] drives the experimental objective and has the intricate knowledge about the subject area
]
]

---

.f-subheadline.hidden[Meet the cast starring today]

The .monash-red2[**"analyst"**] analyses the data after the data is collected.

]
.w-33[

The **"statistician"**

The **"analyst"**
]
.w-33.tl[

The .monash-ruby2[**"statistican"**] creates the experimental design layout after taking into account the statistical and practical constraints.

]
]

---

.f-subheadline.hidden[Meet the cast starring today]

]
.w-33[

The .monash-green2[**"technician"**] carries out the experiment and collects the data.
]
.w-33[
 
<img src="images/stickman/technician2.svg" width="100px" class="img-hor">

The **"technician"**
]
]

---

The statistical software to design experiments
]
.w-33.pr3.pl3[

Good old pen and paper

]
.w-33.pr3.pl3[

The software to enter and store data

]]
 
.gray[Missing prop: the statistical software to analyse experimental data]

---

.center.b[
.f1[The actors are purely illustrative.] 
]

]
.w-70[
.f2[
In practice:

* multiple people can take on each role, 
* one person can take on multiple roles, and/or 
* a person in the role may not specialise in that role (e.g. the statistician role can be acted out by a non-statistician).

]

]
.w-10[

]]

.center[
<img src="images/stickman/technician1.svg" width="100px" class="">
<img src="images/stickman/actor1.svg" width="75px" class="">
<img src="images/stickman/actor2.svg" width="100px" class="">
]

---

---

.flex[
.w-20[
<img src="images/stickman/domain1.svg" width="200px" class="">
]
.w-60[
.sm-quote.pr5.monash-blue2[
Hey, I need to run an experiment. 
Can you test irrigation and fertilizer effect on plant growth?
Bla bla bla...
]

.sm-quote.pl5.monash-ruby2[
{{content}}
]
]
.w-20[
<img src="images/stickman/statistician1.svg" width="200px" class="img-hor">

Okay. **I got the experimental structure *perfectly*.**  I'll go generate the experimental design layout.

---

.flex[
.w-20[
<img src="images/stickman/statistician1.svg" width="200px" class="">
.center[

]
]
.w-60[
.sm-quote.pr5.monash-ruby2[
I have a **_complete_ understanding of the experimental structure** so I shall enter it in the software to generate the experimental design
]

]
.w-20[
 

]]

---

.sm-quote.pl5.monash-green2[
I'll execute this experiment **exactly as planned** and enter the data with **absolutely no mistake**
]

]

]
.w-20[
<img src="images/stickman/technician2.svg" width="150px" class="img-hor">

]]

---

.blockquote[
communication is complex, fraught with tensions, misunderstandings, and problems &mdash; rather than a simple process of creating shared meaning

.footnote.tl.f4[
Littlejohn, Stephen W., Karen A. Foss, and John G. Oetzel. 2017. Theories of Human Communication. Waveland Press, Inc.
]

---

---

.f2.pl3[
* Misunderstandings or incomplete understandings 
]

.sm-quote.pl5.monash-ruby2[
{{content}}
]
]
.w-20[
<img src="images/stickman/statistician1.svg" width="200px" class="img-hor">

Okay. I think I got it. I'll go see what we can do for the experimental design.

---

.f2.pl3[
* A lot of back-and-forth to be on the same page .f3[(which is completely natural and okay)
]]

.flex[
.w-20[
<img src="images/stickman/statistician1.svg" width="200px" class="">
.center[

]
]
.w-60[
.sm-quote.pr5.monash-ruby2[
Okay how about this plan.
]
.center[

]

]
.w-20[
<img src="images/stickman/domain1.svg" width="200px" class="img-hor">
]]

<div class="sm-quote pl5 monash-blue2">
Actually, we can't apply different irrigation methods to these set of plots so we have to change this part. Bla bla.
</div>

<div class="sm-quote pr5 monash-ruby2">
Okay, I'll write this in my notebook.
</div>

---

.f2.pl3[
* OR limited communication and decisions made in silo
]

.flex[
.w-33.br.pa3[
.pull-left[
<img src="images/stickman/domain1.svg" width="120px" class="">
] .sm-quote.monash-blue2[
Oh actually, we didn't have enough seedlings for the test line. Oh well, we'll just leave that empty when we get the plan.
]

]
.w-33.br.pa3[
.pull-left[
<img src="images/stickman/statistician1.svg" width="150px" class="">
] .sm-quote.monash-ruby2[
Hmm, I'm not sure about this but I guess this will do for the experimental design.
]
.center[

]
]
.w-33.pa3[
.pull-left[
<img src="images/stickman/technician1.svg" width="120px" class="">
] .sm-quote.monash-green2[
Oh no! One side of the green house gets more sunlight than the other! Let's move the pots around so that they get the same amount of sunlight.
]
]]

---

.f2.pl3[
* Implicit decisions never explicitly transcribed
]

.flex[
.w-20[
<img src="images/stickman/statistician2.svg" width="100px" class="">
]
.w-60[
.sm-quote.monash-red2.pr5[
Why is the experimental design like this?
]

.sm-quote.monash-ruby2.pl5[
I think I had a good reason at the time but I can't remember!
]

.sm-quote.monash-red2.pr5.f5[
Why are we always on a handstand anyway?
]

.sm-quote.monash-ruby2.pl5.f5[
There's no good reason for that!
]
]
.w-20[
<img src="images/stickman/statistician1.svg" width="120px" class="img-hor">
]]

---

.f2.pl3[
* Knowledge lost
]

.flex[
.w-20[
<img src="images/stickman/statistician2.svg" width="100px" class="">
]
.w-60[
.sm-quote.monash-red2.pr5[
How was the experimental design constructed?
]

.sm-quote.monash-blue2.pl5[
The statistician left a bunch of notes before leaving us for another position but I don't know what's what, so I don't know.
]

]
.w-20[
<img src="images/stickman/domain1.svg" width="100px" class="img-hor">
]]

---

.f2.pl3[
* Thinking analysis will save you 
]

.sm-quote.monash-blue2.pl5[
I think we may have data quality issues in this experimental data.
]

.sm-quote.monash-red2.pr5[
Oh cool data! I can try this fancy statistical model! Hmm, what did you say?
]

.sm-quote.monash-blue2.pl5[
I'm saying that the data may be rubbish.
]

.sm-quote.monash-red2.pr5[
But I got some numbers from fitting this model.
]

]
.w-20[
<img src="images/stickman/domain1.svg" width="100px" class="img-hor">
]]

---

# Garbage in, garbage out

.w-80.mauto.f1[
No statistical model, however complex it is, can make any sense of the data if the collected data is rubbish

]

---

# Redoing an experiment is *expensive*
--

.f1[
There's a higher stake in getting the experiment design wrong compared to getting the analysis wrong.
]

In some cases, redoing an experiment is not every possible!

---

# What most statistics (and data science) research is about

.footnote.f5[
Grolemund, G., & Wickham, H. (2017). R for Data Science. O'Reilly Media.
]

---

# World of data collection

But there is a whole world of data collection before importing data!

* There is probably more gain in extracting information in ensuring the quality of the data than analysis.
--

* Experimental design research is generally concerned about generating the experimental design layout
--

* `edibble` aims to complement many great experimental design research to design the whole experiment

---

# Overview of `edibble`

---

# The grammar of experimental designs

<img src="images/edibble.png" style="width:200px" class="pull-right"> .info-box.w-60[
An abstract computational framework that maps fundamental experimental components to an object oriented system to build and modify experimental designs. 
Currently implemented as the `edibble` R-package.
]

.f2[
 **Package documentation**:
.center[
<a href='https://edibble.emitanaka.org'>edibble.emitanaka.org</a>
]

**Source code**:

**Name origin**: produce **e**xperimental **d**esign table (or `t**ibble**`)

]

.footnote.f5[
Kirill Müller and Hadley Wickham (2021). tibble: Simple Data Frames. R package version 3.1.5. https://CRAN.R-project.org/package=tibble

]

---

# Lifecycle

]

.f2[
* Currently `edibble` is <a href="https://lifecycle.r-lib.org/articles/stages.html#experimental"><img src="images/lifecycle-experimental.svg" style = "height:1em"></a>
* Some functions like `allocation_trts` and `randomise_trts` have become <a href="https://lifecycle.r-lib.org/articles/stages.html#deprecated"><img src="images/lifecycle-deprecated.svg" style = "height:1em"></a> in favour of `allot_trts` and `assign_trts`
]

.footnote.f5[
Lionel Henry and Hadley Wickham (2021). lifecycle:
  Manage the Life Cycle of your Package Functions. R
  package version 1.0.1.
  https://CRAN.R-project.org/package=lifecycle
]

---

# Reframing how you think about experimental designs

.blockquote.w-100[
.pl3[
Good design considers units and treatments first, and then allocates treatments to units. It does not choose from a menu of named designs.

.footnote.f5[
Bailey, R. (2008). Design of Comparative Experiments (Cambridge Series in Statistical and Probabilistic Mathematics). Cambridge: Cambridge University Press. doi:10.1017/CBO9780511611483
]

]

.f2[
* Nevertheless named experimental designs are very prevelant and can be useful to describe particular designs succintly!]

---

# Example .circle[1] Completely randomised design

]
.w-70[

```r
library(edibble)
*code_classical("crd", t = 3, n = 6)
```

```
set.seed(648)
start_design("crd") %>%
  set_units(unit = 6) %>%
  set_trts(treat = 3) %>%
  allot_trts(treat ~ unit) %>%
  assign_trts("random") %>%
  serve_table()
```

* Note: currently only limited named experimental designs are supported
* this function name and arguments will likely change in near future
]

]

---

# Example .circle[2] Randomised complete block design
.flex[
.w-30[

]
.w-70[

```r
library(edibble)
*code_classical("rcbd", t = 3, b = 2)
```

```
set.seed(231)
start_design("rcbd") %>%
  set_units(block = 2,
            unit = nested_in(block, 3)) %>%
  set_trts(treat = 3) %>%
  allot_trts(treat ~ unit) %>%
  assign_trts("random") %>%
  serve_table()
```

* this function name and arguments will likely change in near future
]
]

---

# Example .circle[3] Factorial design
.flex[
.w-30[

]
.w-70[

```r
library(edibble)
*code_classical("factorial", trt = c(2, 4), n = 16)
```

```
set.seed(289)
start_design("factorial") %>%
  set_units(unit = 16) %>%
  set_trts(treat1 = 2,
           treat2 = 4) %>%
  allot_trts( ~ unit) %>%
  assign_trts("random") %>%
  serve_table()
```

* this function name and arguments will likely change in near future
]
]

---

# Example .circle[4] Split-plot design
.flex[
.w-30[

]
.w-70[

```r
library(edibble)
*code_classical("split", t1 = 2, t2 = 4, r = 2)
```

```
set.seed(566)
start_design("split") %>%
  set_units(mainplot = 4,
            subplot = nested_in(mainplot, 4)) %>%
  set_trts(treat1 = 2,
           treat2 = 4) %>%
  allot_trts(treat1 ~ mainplot,
                treat2 ~ subplot) %>%
  assign_trts("random") %>%
  serve_table()
```

* this function name and arguments will likely change in near future
]
]

---

# Example .circle[5] Strip-plot design
.flex[
.w-30[

]
.w-70[

]
]

```r
library(edibble)
start_design("Strip-plot") %>% 
  set_trts(diet = 4,
           breed = 5) %>% 
  set_units(hen = 5,
            order = 4,
            chick = ~hen:order) %>% 
  allot_trts(breed ~ hen,
             diet ~ order) %>% 
  assign_trts("random") %>% 
  serve_table()
```

```
## # An edibble: 20 x 5
## diet breed hen order chick
## <trt(4)> <trt(5)> <unit(5)> <unit(4)> <unit(20)>
## 1 diet1 breed1 hen1 order1 chick1 
## 2 diet1 breed4 hen2 order1 chick2 
## 3 diet1 breed2 hen3 order1 chick3 
## 4 diet1 breed3 hen4 order1 chick4 
## 5 diet1 breed5 hen5 order1 chick5 
## 6 diet3 breed1 hen1 order2 chick6 
## 7 diet3 breed4 hen2 order2 chick7 
## 8 diet3 breed2 hen3 order2 chick8 
## 9 diet3 breed3 hen4 order2 chick9 
## 10 diet3 breed5 hen5 order2 chick10
## 11 diet2 breed1 hen1 order3 chick11
## 12 diet2 breed4 hen2 order3 chick12
## 13 diet2 breed2 hen3 order3 chick13
## 14 diet2 breed3 hen4 order3 chick14
## 15 diet2 breed5 hen5 order3 chick15
## 16 diet4 breed1 hen1 order4 chick16
## 17 diet4 breed4 hen2 order4 chick17
## 18 diet4 breed2 hen3 order4 chick18
## 19 diet4 breed3 hen4 order4 chick19
## 20 diet4 breed5 hen5 order4 chick20
```

---

# Experimental context is important

* Name the variables so it always reminds you of the context

.flex.f4[
.w-50.br[

```r
start_design("My plant experiment") %>%
  set_units(mainplot = 4,
            subplot = nested_in(mainplot, 4)) %>%
  set_trts(water = c("irrigated", "rain-fed"),
           variety = 4) %>%
  allot_trts(water ~ mainplot,
             variety ~ subplot) %>%
  assign_trts("random", seed = 1) %>%
  serve_table()
```

```
## # An edibble: 16 x 4
## mainplot subplot water variety
## <unit(4)> <unit(16)> <trt(2)> <trt(4)>
## 1 mainplot1 subplot1 irrigated variety3
## 2 mainplot1 subplot2 irrigated variety1
## 3 mainplot1 subplot3 irrigated variety4
## 4 mainplot1 subplot4 irrigated variety2
## 5 mainplot2 subplot5 irrigated variety2
## 6 mainplot2 subplot6 irrigated variety4
## 7 mainplot2 subplot7 irrigated variety3
## 8 mainplot2 subplot8 irrigated variety1
## 9 mainplot3 subplot9 rain-fed variety3
## 10 mainplot3 subplot10 rain-fed variety1
## 11 mainplot3 subplot11 rain-fed variety4
## 12 mainplot3 subplot12 rain-fed variety2
## 13 mainplot4 subplot13 rain-fed variety1
## 14 mainplot4 subplot14 rain-fed variety4
## 15 mainplot4 subplot15 rain-fed variety3
## 16 mainplot4 subplot16 rain-fed variety2
```
]
.w-50[

```r
start_design("My animal experiment") %>%
  set_units(pen = 4,
            cow = nested_in(pen, 4)) %>%
  set_trts(diet = c("low-card", "high-fat"),
           breed = 4) %>%
  allot_trts(diet ~ pen,
             breed ~ cow) %>%
  assign_trts("random", seed = 1) %>%
  serve_table()
```

```
## # An edibble: 16 x 4
## pen cow diet breed
## <unit(4)> <unit(16)> <trt(2)> <trt(4)>
## 1 pen1 cow1 low-card breed3
## 2 pen1 cow2 low-card breed1
## 3 pen1 cow3 low-card breed4
## 4 pen1 cow4 low-card breed2
## 5 pen2 cow5 low-card breed2
## 6 pen2 cow6 low-card breed4
## 7 pen2 cow7 low-card breed3
## 8 pen2 cow8 low-card breed1
## 9 pen3 cow9 high-fat breed3
## 10 pen3 cow10 high-fat breed1
## 11 pen3 cow11 high-fat breed4
## 12 pen3 cow12 high-fat breed2
## 13 pen4 cow13 high-fat breed1
## 14 pen4 cow14 high-fat breed4
## 15 pen4 cow15 high-fat breed3
## 16 pen4 cow16 high-fat breed2
```

]]

---

# Example .circle[6] Calf feeding experiment

]
.w-70.f4[

]
]

```r
start_design("Calf feeding") %>% 
* set_context(location = "Wagga Wagga") %>%
  set_trts(feed = 3) %>% 
  set_units(pen = 6, 
            calf = nested_in(pen, 4)) %>% 
  allot_trts(feed ~ pen) %>% 
  assign_trts("random") %>% 
* set_rcrds_of(calf = c("milk", "weight")) %>%
* expect_rcrds(milk = to_be_numeric(with_value(">=", 0)),
*              yield = to_be_numeric(with_value(">=", 0))) %>%
  serve_table() %>% 
  export_design("calf-design.xlsx", overwrite = TRUE)
```

```
## Loading required package: openxlsx
```

```
## Wagga Wagga
```

```
## ✓ Calf feeding has been written to 'calf-design.xlsx'
```

---

# Example .circle[7] Chick weight

]
.w-50[

]]

.footnote.f5[
Experiment based from Martin et al. (1996) Effects of grasshopper-control insecticides on survival and brain acetylcholinesterase of pheasant (Phasianus colchicus) chicks.
Environmental Toxicology and Chemistry **15**(4) 518-524.
]

* Three insecticides were used, at low and high doses.
{{content}}
--

* The low dose was the highest dose recommended by the department of agriculture; the high dose was four times as much as the recommended dose, to assess the effects
of mistakes.
{{content}}
--

* The experimental procedure took place in each of three consecutive weeks.
{{content}}
--

* On the first day of each week a number of newly-hatched female pheasant chicks
were placed in a brooder pen.
{{content}}
--

* On the third day, the chicks were randomly divided into twelve groups of six chicks
each.
{{content}}
--

* Each chick was given an identification tape and weighed.
{{content}}
--

* On the fourth day, a portion of the field was divided into three strips, each of which
was divided into two swathes.
{{content}}
--

* The two swathes within each strip were sprayed with the two doses of the same
insecticide.

---

# Example .circle[7] Chick weight

.flex.f5[
.w-50[
* An experiment was conducted on a prairie in Western Canada to find out if insecticides used to control grasshoppers affected the weight of young chicks of ring-necked pheasants, either by affecting the grass around the chicks or by affecting the
grasshoppers eaten by the chicks.
* Three insecticides were used, at low and high doses.
* The low dose was the highest dose recommended by the department of agriculture;
the high dose was four times as much as the recommended dose, to assess the effects
of mistakes.
* The experimental procedure took place in each of three consecutive weeks.
* On the first day of each week a number of newly-hatched female pheasant chicks
were placed in a brooder pen.
* On the third day, the chicks were randomly divided into twelve groups of six chicks
each.
* Each chick was given an identification tape and weighed.
* On the fourth day, a portion of the field was divided into three strips, each of which
was divided into two swathes.
* The two swathes within each strip were sprayed with the two doses of the same
insecticide.

]
.w-50[
* Two pens were erected on each swathe, and one group of pheasant chicks was put
into each pen.
{{content}}

]]

* For the next 48 hours, the chicks were fed with grasshoppers which had been collected
locally.
{{content}}
--

* Half the grasshoppers were anaesthetized and sprayed with insecticide; the other half
were also anaesthetized and handled in every way like the first half except that they
were not sprayed.
{{content}}
--

* All grasshoppers were frozen.
{{content}}
--

* The experimenters maintained a supply of frozen grasshoppers to each pen, putting
them on small platforms so that they would not absorb further insecticide from the
grass.
{{content}}
--

* In each swathe, one pen had unsprayed grasshoppers while the other had grasshoppers sprayed by the insecticide which had been applied to that swathe.
{{content}}
--

* At the end of the 48 hours, the chicks were weighed again individually. 
{{content}}
--

---

# Example .circle[7] Chick weight: the `edibble` code

```r
start_design("Chick weight") %>% 
  set_trts(insecticide = 3,
           dose = c("low", "high")) %>% 
  set_units(week = 3,
            strip = nested_in(week, 4),
            swathes = nested_in(strip, 2),
            pen = nested_in(swathes, 2),
            chick = nested_in(pen, 6)) %>% 
  set_trts(food = c("spray", "no-spray")) %>% 
  allot_trts(insecticide ~ strip,
             dose ~ swathes,
             food ~ pen) %>% 
  assign_trts("random") %>% 
  set_rcrds(weight = chick) %>% 
  serve_table()
```

```
## # An edibble: 288 x 9
## insecticide dose week strip swathes pen
## <trt(3)> <trt(2)> <unit(3)> <unit(12)> <unit(24)> <unit(48)>
## 1 insecticide2 high week1 strip1 swathes1 pen1
## 2 insecticide2 high week1 strip1 swathes1 pen1
## 3 insecticide2 high week1 strip1 swathes1 pen1
## 4 insecticide2 high week1 strip1 swathes1 pen1
## 5 insecticide2 high week1 strip1 swathes1 pen1
## 6 insecticide2 high week1 strip1 swathes1 pen1
## 7 insecticide2 high week1 strip1 swathes1 pen2
## 8 insecticide2 high week1 strip1 swathes1 pen2
## 9 insecticide2 high week1 strip1 swathes1 pen2
## 10 insecticide2 high week1 strip1 swathes1 pen2
## # … with 278 more rows, and 3 more variables: chick <unit(288)>, food <trt(2)>,
## # weight <rcrd>
```

]

---
# Example .circle[7] Chick weight: closer look

.flex[
.w-50[
.overflow-scroll.vh-100[
<img src="images/eg4-chicks.png" width="100%">
]
.center[
scroll down to see the rest of image
]
]
.w-50.pl3[
<img src="images/eg4-chicks-anova.png" width="100%">
]]

---

# Example .circle[8] Unbalanced designs

.f2[
All designs thus far have been **balanced** (i.e. equal replicate) and **complete** (each treatment appears the same number of times in each block)... 
]

```r
start_design("unbalanced & incomplete") %>% 
  set_units(site = c("Horsham", "Narrabri", 
                     "Wagga", "Roseworthy"),
*           plot = nested_in(site,
*                            "Horsham" ~ 6,
*                            "Narrabri" ~ 3,
*                            . ~ 4)) %>%
  set_trts(breed = 4) %>% 
  allot_trts(breed ~ plot) %>% 
  assign_trts("random") %>% 
  serve_table()
```

```
## # An edibble: 17 x 3
## site plot breed
## <unit(4)> <unit(17)> <trt(4)>
## 1 Horsham plot1 breed1
## 2 Horsham plot2 breed1
## 3 Horsham plot3 breed4
## 4 Horsham plot4 breed2
## 5 Horsham plot5 breed3
## 6 Horsham plot6 breed4
## 7 Narrabri plot7 breed1
## 8 Narrabri plot8 breed3
## 9 Narrabri plot9 breed2
## 10 Wagga plot10 breed4
## 11 Wagga plot11 breed1
## 12 Wagga plot12 breed3
## 13 Wagga plot13 breed2
## 14 Roseworthy plot14 breed2
## 15 Roseworthy plot15 breed3
## 16 Roseworthy plot16 breed4
## 17 Roseworthy plot17 breed1
```
]
.w-50.bl.pl3.f4[
Reference level by number:

```r
start_design("unbalanced & incomplete") %>% 
  set_units(site = c("Horsham", "Narrabri", 
                     "Wagga", "Roseworthy"),
*           plot = nested_in(site,
*                            1 ~ 6,
*                            2 ~ 3,
*                            . ~ 4)) %>%
  set_trts(breed = 4) %>% 
  allot_trts(breed ~ plot) %>% 
  assign_trts("random") %>% 
  serve_table()
```

```
## # An edibble: 17 x 3
## site plot breed
## <unit(4)> <unit(17)> <trt(4)>
## 1 Horsham plot1 breed4
## 2 Horsham plot2 breed4
## 3 Horsham plot3 breed2
## 4 Horsham plot4 breed1
## 5 Horsham plot5 breed3
## 6 Horsham plot6 breed2
## 7 Narrabri plot7 breed2
## 8 Narrabri plot8 breed3
## 9 Narrabri plot9 breed1
## 10 Wagga plot10 breed3
## 11 Wagga plot11 breed1
## 12 Wagga plot12 breed4
## 13 Wagga plot13 breed2
## 14 Roseworthy plot14 breed2
## 15 Roseworthy plot15 breed1
## 16 Roseworthy plot16 breed4
## 17 Roseworthy plot17 breed3
```
]]

---

background-image: url("images/workflow.png")
background-size: 80%
.absolute.f5.right-0[
.blockquote.w-60[
communication is complex, fraught with tensions, misunderstandings, and problems &mdash; rather than a simple process of creating shared meaning

<img src="images/stickman/technician2.svg" width="100px" class="">
]

.absolute.right-3[
<img src="images/stickman/statistician2.svg" width="100px" class="img-hor">
]

.absolute.bottom-0.right-1[
.f1[What can help to communicate more effectively?]
]

---

# Grammar of graphics with `ggplot2`

---

# Grammar of graphics: origin and implementations

* Initial instances of the grammar of graphics was mentioned by William C. Brinton
* A full computational framework was developed by Leland Wilkinson with implementation in **SYSTAT**.

.footnote.f4[
Brinton (1914) Graphic methods for presenting facts 
Wilkinson (1999) The Grammar of Graphics. *Statistics and Computing. Springer, 1st edition.*
]

* An interpretation of the grammar of graphics by Hadley Wickham (as part of his PhD, 2008) was implemented in **R** as the `ggplot2` package.
--

* Emulation of `ggplot2` in **python** started to be developed: 
   * `ggpy` (defunct) 
   * `plotnine` by Hassan Kibirige,
   * `seaborn` by Michael Waskom .f7[(this one is not quite trying to emulate ggplot)]
--

* In **Julia**, `Gadfly` by Daniel C. Jones implements the grammar of graphics. 
--

* In **Matlab**, `gramm` by Pierre Morel implements the grammar as a toolbox.
--

* In **Javascript**, `G2` by AntV team, adding also interactivity, with this version emulated in R as `g2r` package.
--

* `ggplot2` is arguably the most popular interpretation of grammar of graphics with over 35,000 citations

---

# .square[Data .circle[1]] Crop production in Australia

.measure.f3.pa3[
In Australia, total production of each crop nationally are

<table class=" lightable-classic" style='font-family: "Arial Narrow", "Source Sans Pro", sans-serif; width: auto !important; margin-left: auto; margin-right: auto;'>
 <thead>
 <tr>
 <th style="text-align:left;"> Crop </th>
 <th style="text-align:right;"> Production ('000t) </th>
 </tr>
 </thead>
<tbody>
 <tr>
 <td style="text-align:left;"> barley </td>
 <td style="text-align:right;"> 13,414 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> sorghum </td>
 <td style="text-align:right;"> 1,208 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> maize </td>
 <td style="text-align:right;"> 467 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> oats </td>
 <td style="text-align:right;"> 1,879 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> triticale </td>
 <td style="text-align:right;"> 247 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> wheat </td>
 <td style="text-align:right;"> 35,134 </td>
 </tr>
</tbody>
</table>
]

.footnote.f4[
Data source: ABARES Australian Crop Report, 2017, No. 181
]

---
name: base-plot

# Plotting with "base R"

.flex.h-80[
.w33.br[

```r
df
```

```
##        crop  prod
## 1    barley 13414
## 2   sorghum  1208
## 3     maize   467
## 4      oats  1879
## 5 triticale   247
## 6     wheat 35134
```
]
.w33.br[

```r
barplot(as.matrix(df$prod),
        legend = df$crop)
```

]
.w33[

```r
pie(df$prod, labels = df$crop)
```

]
]

.footnote.f4[
 R Core Team (2020) R: A Language and Environment for Statistical Computing https://www.R-project.org/
]

.absolute.bottom-3.left-1[
.note-box.w-80[
Single purpose functions to generate "named plots"
]]

---

# Plotting with the `ggplot2` R-package

.flex.h-80[
.w-33.br[

```r
df
```

```
##        crop  prod
## 1    barley 13414
## 2   sorghum  1208
## 3     maize   467
## 4      oats  1879
## 5 triticale   247
## 6     wheat 35134
```
]
.w-33.br.f4[

```r
ggplot(df, aes(x = "", # dummy
               y = prod, 
               fill = crop)) + 
  geom_col(color = "black")
```

]
.w-33.f4[

]
]

```r
ggplot(df, aes(x = "", # dummy
               y = prod, 
               fill = crop)) + 
  geom_col(color = "black") + 
* coord_polar(theta = "y")
```

.absolute.bottom-2.left-1[
.info-box.w-60[
The difference between a **stacked barplot** and a **pie chart** is that the coordinate system is transformed from the **Cartesian coordinate** to **polar coordinate**. 
]]

---

# .square[Data .circle[2]] Potato scab infection with sulfur treatments

Let's look at another data...

.flex.f4[
.w-33[
.overflow-scroll.h---400[

```r
data(cochran.crd, package = "agridat")
cochran.crd
```

```
##    inf trt row col
## 1    9  F3   4   1
## 2   12   O   4   2
## 3   18  S6   4   3
## 4   10 F12   4   4
## 5   24  S6   4   5
## 6   17 S12   4   6
## 7   30  S3   4   7
## 8   16  F6   4   8
## 9   10   O   3   1
## 10   7  S3   3   2
## 11   4 F12   3   3
## 12  10  F6   3   4
## 13  21  S3   3   5
## 14  24   O   3   6
## 15  29   O   3   7
## 16  12  S6   3   8
## 17   9  F3   2   1
## 18   7 S12   2   2
## 19  18  F6   2   3
## 20  30   O   2   4
## 21  18  F6   2   5
## 22  16 S12   2   6
## 23  16  F3   2   7
## 24   4 F12   2   8
## 25   9  S3   1   1
## 26  18   O   1   2
## 27  17 S12   1   3
## 28  19  S6   1   4
## 29  32   O   1   5
## 30   5 F12   1   6
## 31  26   O   1   7
## 32   4  F3   1   8
```
]
]
.w-33[

```r
ggplot(cochran.crd, 
       aes(col, row, 
           fill = trt)) +
  geom_tile(color = "black", 
            size = 1.3)
```

]
.w-33[

```r
ggplot(cochran.crd, 
       aes(col, row, 
           fill = inf)) +
  geom_tile(color = "black", 
            size = 1.3)
```

]

---

# Making publication ready plots with `ggplot2`

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) 
```
]
.w-40[
<img src="images/cochran-1-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") 
```
]
.w-40[
<img src="images/cochran-2-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") +
 labs(title = "Potato scab infection with sulfur\ntreatments")
```
]
.w-40[
<img src="images/cochran-3-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") +
 labs(title = "Potato scab infection with sulfur\ntreatments",
 x = "Column", y = "Row")
```
]
.w-40[
<img src="images/cochran-4-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") +
 labs(title = "Potato scab infection with sulfur\ntreatments",
 x = "Column", y = "Row",
 caption = "Data source: Tamura, R.N. and Nelson, L.A. and Naderman, G.C., (1988).\nAn investigation of the validity and usefulness of trend analysis for field plot data.\nAgronomy Journal, 80, 712-718.")
```
]
.w-40[
<img src="images/cochran-5-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") +
 labs(title = "Potato scab infection with sulfur\ntreatments",
 x = "Column", y = "Row",
 caption = "Data source: Tamura, R.N. and Nelson, L.A. and Naderman, G.C., (1988).\nAn investigation of the validity and usefulness of trend analysis for field plot data.\nAgronomy Journal, 80, 712-718.") + 
 theme(text = element_text(size = 20))
```
]
.w-40[
<img src="images/cochran-6-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") +
 labs(title = "Potato scab infection with sulfur\ntreatments",
 x = "Column", y = "Row",
 caption = "Data source: Tamura, R.N. and Nelson, L.A. and Naderman, G.C., (1988).\nAn investigation of the validity and usefulness of trend analysis for field plot data.\nAgronomy Journal, 80, 712-718.") + 
 theme(text = element_text(size = 20),
 plot.caption = element_text(size = 12))
```
]
.w-40[
<img src="images/cochran-7-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") +
 labs(title = "Potato scab infection with sulfur\ntreatments",
 x = "Column", y = "Row",
 caption = "Data source: Tamura, R.N. and Nelson, L.A. and Naderman, G.C., (1988).\nAn investigation of the validity and usefulness of trend analysis for field plot data.\nAgronomy Journal, 80, 712-718.") + 
 theme(text = element_text(size = 20),
 plot.caption = element_text(size = 12),
 plot.title.position = "plot",
 plot.caption.position = "plot",)
```
]
.w-40[
<img src="images/cochran-8-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

# Making publication ready plots with `ggplot2`

```r
ggplot(cochran.crd, aes(col, row, fill = inf)) +
 geom_tile(color = "black", size = 1.3) + 
 scale_fill_gradient(low = "white", high = "firebrick") + 
 labs(title = "Potato scab infection with sulfur\ntreatments",
 x = "Column", y = "Row",
 caption = "Data source: Tamura, R.N. and Nelson, L.A. and Naderman, G.C., (1988).\nAn investigation of the validity and usefulness of trend analysis for field plot data.\nAgronomy Journal, 80, 712-718.") + 
 theme(text = element_text(size = 20),
 plot.caption = element_text(size = 12),
 plot.title.position = "plot",
 plot.caption.position = "plot",
 panel.background = element_blank())
```
]
.w-40[
<img src="images/cochran-final-1.png" width="432" style="display: block; margin: auto;" />
]
]

---

# Many extension packages exists for `ggplot2`

```r
ggplot(cochran.crd, aes(trt, inf, color = trt)) +
  geom_boxplot() +
  theme_bw(base_size = 18)
```

---

# Many extension packages exists for `ggplot2`

```r
ggplot(cochran.crd, aes(trt, inf, color = trt)) +
* ggbeeswarm::geom_quasirandom() +
  theme_bw(base_size = 18)
```

---

# Visualising experimental designs with the `deggust` R-package

---

# Visualising experimental designs

<img src="images/deggust.png" style="width:200px" class="pull-right"> .info-box.w-60[
The `deggust` R-package aims to convert `edibble` designs to `ggplot` objects *seamlessly*.
]

Currently under developed!

.f2[
 **Package documentation**:
.center[
<a href='https://deggust.emitanaka.org'>deggust.emitanaka.org</a>
]

**Source code**:

**Name origin**: deggust as in *degust*, and &emsp;&nbsp;&nbsp; make **d**esign of **e**xperiments into `**gg**plot` objects

]

.footnote.f5[
H. Wickham. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York, 2016.

]

---

# Example: Pig diet experiment

```r
library(edibble)
plan <- start_design("Pig diet experiment") %>% 
 set_trts(diet = c("carb", "protein", "fat")) %>% 
 set_units(pig = 50) %>% 
 allot_trts(diet ~ pig) %>% 
 assign_trts("random", seed = 1) %>% 
 serve_table()

plan
```

```
## # An edibble: 50 x 2
## diet pig
## <trt(3)> <unit(50)>
## 1 carb pig1 
## 2 carb pig2 
## 3 protein pig3 
## 4 carb pig4 
## 5 protein pig5 
## 6 fat pig6 
## 7 fat pig7 
## 8 fat pig8 
## 9 protein pig9 
## 10 carb pig10
## # … with 40 more rows
```

---

# Visualising designs with `ggplot2`

```r
library(ggplot2)
plan %>% 
  edibble::as_data_frame() %>% # in the future this step will not be needed
  ggplot(aes(pig, "1", fill = diet)) +
  geom_tile(color = "black")
```

* Slightly painful if you want to *quickly* visualise your design.

* Also not a great visualisation

---

# `deggust::autoplot()` .f3[Part 1]

Just `autoplot` it!

```r
library(deggust)
```

```r
autoplot(plan)
```

---

# `deggust::autoplot()` .f4[Part 2]

* It makes some decision for you of how to plot which can be customised in two ways:
   1. modify scale and theme like any ggplot objects!
   2. as arguments in the `autoplot` function

]
.w-50[

]

---

# `deggust::autoplot()` .f4[Part 3]

* It makes some decision for you of how to plot which can be customised in two ways:
   1. .red[**modify scale and theme like any ggplot objects!**]
   2. as arguments in the `autoplot` function

```r
autoplot(plan) +
 # ggplot2 functions below
* theme_void() +
* scale_fill_viridis_d(option = "A")
```
]
.w-50[
<img src="images/vis-autoplot-modify-1.png" width="432" style="display: block; margin: auto;" />
]

]

---

# `deggust::autoplot()` .f4[Part 4]

* It makes some decision for you of how to plot which can be customised in two ways:
   1. modify scale and theme like any ggplot objects!
   2. .red[**as arguments in the `autoplot` function**]

```r
autoplot(plan, 
* shape = "hexagon",
* text = TRUE,
* aspect_ratio = 0.5)
```
]
.w-50[
<img src="images/vis-autoplot-args-1.png" width="432" style="display: block; margin: auto;" />
]

]

---

# `deggust::autoplot()` .f4[Part 5]

<img src="images/pig.png" width = "200px">
--

```r
autoplot(plan, 
* image = "images/pig.png") +
 theme_void() 
```
]
.w-50[
<img src="images/vis-autoplot-pigimage-1.png" width="432" style="display: block; margin: auto;" />
]

]

---

# `deggust::autoplot()` .f4[Part 6]

* Nested design

```r
start_design() %>% 
 set_units(pen = 10, 
 pig = nested_in(pen, 5)) %>% 
 set_trts(breed = c("A", "B", "C")) %>% 
 allot_trts(breed ~ pig) %>% 
 assign_trts("random", seed = 2021) %>% 
 serve_table() %>% 
 autoplot()
```
]
.w-50[
<img src="images/vis-nested-1.png" width="432" style="display: block; margin: auto;" />

]

---

# `deggust::autoplot()` .f4[Part 7]

* What changed here?

```r
start_design() %>% 
 set_units(pen = 10, 
 pig = nested_in(pen, 5)) %>% 
 set_trts(breed = c("A", "B", "C")) %>% 
 allot_trts(breed ~ pen) %>% 
 assign_trts("random", seed = 2021) %>% 
 serve_table() %>% 
 autoplot()
```
]
.w-50[
<img src="images/vis-nested-pen-1.png" width="432" style="display: block; margin: auto;" />

]

---

# `deggust::autoplot()` .f4[Part 8]

* Factorial experiment

* Note: scale will be fixed so it's easier to distinguish between different treatment factors
]
.w-50[
<img src="images/vis-nested-fac-1.png" width="432" style="display: block; margin: auto;" />

]

---

# `deggust::autoplot()` .f4[Part 9]

* Focus on the treatment you want:

```r
start_design() %>% 
 set_units(pig = 40) %>% 
 set_trts(breed = c("A", "B", "C"),
 feed = c("X", "Y", "Z")) %>% 
 allot_trts(breed:feed ~ pig) %>% 
 assign_trts("random", seed = 2021) %>% 
 serve_table() %>% 
* autoplot(trts = "breed")
```
]
.w-50[
<img src="images/vis-nested-fac1-1.png" width="432" style="display: block; margin: auto;" />

]

---

# `deggust::autoplot()` .f4[Part 10]

* Is your design too big to fit in the plot?

```r
start_design() %>% 
 set_units(pen = 100,
 pig = nested_in(pen, 10)) %>% 
 set_trts(breed = c("A", "B", "C")) %>% 
 allot_trts(breed ~ pig) %>% 
 assign_trts("random", seed = 2021) %>% 
 serve_table() %>% 
 autoplot()
```
]
.w-50[
<img src="images/vis-nested-big-1.png" width="432" style="display: block; margin: auto;" />

]

---

# `deggust::autoplot()` .f4[Part 10]

* Is your design too big to fit in the plot?
* Subset it!

```r
start_design() %>% 
 set_units(pen = 100,
 pig = nested_in(pen, 10)) %>% 
 set_trts(breed = c("A", "B", "C")) %>% 
 allot_trts(breed ~ pig) %>% 
 assign_trts("random", seed = 2021) %>% 
 serve_table() %>% 
* dplyr::filter(pen %in% c("pen1", "pen2", "pen3", "pen4")) %>%
 autoplot() +
 annotate("text", x = 10, y = 4.7, label = "x 25 more")
```
]
.w-50[
<img src="images/vis-nested-filter-1.png" width="432" style="display: block; margin: auto;" />

]

---

# Summary

.flex[
.w-60[
* The .monash-blue[grammar of experimental designs] is an abstract computational framework that .monash-blue[encourages a higher-order thinking] by enforcing the experimental structure and context
{{content}}
]
.w-40.center[
<img src="images/edibble.png" height="120px"> <img src="images/deggust.png" height="120px">

]]

* `edibble` is designed to be user friendly and accommodate natural order of thinking for specifying experimental structure
{{content}}
--

* The grammar makes each step modular... you can easily extend it (like `deggust`) or mix-and-match methods
{{content}}
--

* This makes it easier to leverage existing functionalities in `edibble` so other developers can focus on what they want to do the most 
{{content}}
--

* And hopefully this framework becomes a common base that promotes collaboration and knowledge sharing

---

# Thanks for listening!

<ul class="fa-ul f2">
 <li>Slides: <a href='https://emitanaka.org/slides/stats4bio2021/deggust'>emitanaka.org/slides/stats4bio2021/deggust</a>
</li>
 <li> <code>edibble</code> package documentation: <a href='https://edibble.emitanaka.org'>edibble.emitanaka.org</a></li>
 <li><code>edibble</code> source code: <a href='https://github.com/emitanaka/edibble'>github.com/emitanaka/edibble</a></li>
 <li> <code>deggust</code> package documentation: <a href='https://deggust.emitanaka.org'>deggust.emitanaka.org</a></li>
 <li><code>deggust</code> source code: <a href='https://github.com/emitanaka/deggust'>github.com/emitanaka/deggust</a></li>
 <li><a href="mailto:emi.tanaka@monash.edu">emi.tanaka@monash.edu</a> <a href="https://twitter.com/statsgen">@statsgen</a></li>
</ul>

.f2[
* Feature requests or issues with `edibble` or `deggust`? Submit or upvote here: <a href='https://github.com/emitanaka/edibble/issues'>github.com/emitanaka/edibble/issues</a>, <a href='https://github.com/emitanaka/deggust/issues'>github.com/emitanaka/deggust/issues</a>, send me an email or tell me now!
]