/
README.Rmd
152 lines (112 loc) · 5.42 KB
/
README.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
---
output: github_document
---
<!-- README.md is generated from README.Rmd. Please edit that file -->
```{r setup, include = FALSE}
knitr::opts_chunk$set(
collapse = TRUE,
comment = "#>",
fig.path = "man/figures/README-",
out.width = "100%",
cache = TRUE,
cache.extra = 42
)
```
# metamer
<!-- badges: start -->
[](https://zenodo.org/badge/latestdoi/159563811) [](https://travis-ci.org/eliocamp/metamer) [](https://app.codecov.io/gh/eliocamp/metamer?branch=master)
<!-- badges: end -->
Implements the algorithm proposed by [Matejka & Fitzmaurice (2017)](https://www.autodesk.com/research/publications/same-stats-different-graphs) to create metamers (datasets with identical statistical properties but very different graphs) with an annealing scheme derived from [de Vicente et al. (2003)](https://www.sciencedirect.com/science/article/abs/pii/S0375960103013653?via%3Dihub).
In colour theory, [metamers](https://en.wikipedia.org/wiki/Metamerism_(color)) are colours that have very different wavelength distribution but are perceived as equal by out visual system. This happens because out eyes essentially summarise a continuous distribution of wavelength by just 3 numbers: the amount that each type of cone cell is exited. Colour metamerism is how artists can reproduce so many colours with a few pigments, or how PC monitors use only 3 lights to show colourful pictures.
(from the excellent [Color: From Hexcodes to Eyeballs](http://jamie-wong.com/post/color/) by [Jamie Wong](https://github.com/jlfwong))
Statistical transformations such as mean, standard deviation and correlation behave very similarly in that they summarise data with just a few numbers for the benefit of our limited cognitive capacity. Thus, statistical metamers are sets of data that share some statistical properties.
[This article](https://eliocamp.github.io/codigo-r/en/2019/01/statistical-metamerism/) explores statistical metamerism in more detail.
## Installation
You can install metamer with:
```r
install.packages("metamer")
```
or install the development version with:
``` r
# install.packages("devtools")
devtools::install_github("eliocamp/metamer")
```
## Example
You can construct metamers from a starting dataset and a vector of statistical properties to remain constant (by default, up to 2 significant figures).
```{r example}
library(metamer)
# Start with the datasaurus
# install.packages("datasauRus")
dino <- subset(datasauRus::datasaurus_dozen, dataset == "dino")
dino$dataset <- NULL
# And we want to preserve means and correlation
mean_cor <- delayed_with(mean(x), mean(y), cor(x, y))
set.seed(42) # To make results reproducible
metamers <- metamerise(dino, preserve = mean_cor,
stop_if = n_metamers(300),
perturbation = 1,
keep = 19)
print(metamers)
```
We found `r length(metamers$metamers)` metamers. Let's see the final one, with the starting dataset as background.
```{r}
library(ggplot2)
ggplot(tail(metamers), aes(x, y)) +
geom_point(data = dino, color = "red", alpha = 0.5, size = 0.4) +
geom_point()
```
We can check that the statistical properties have been preserved up to 2 significant figures:
```{r}
cbind(dino = signif(mean_cor(dino), 2),
last = signif(mean_cor(tail(metamers)), 2))
```
However, a semi random cloud of points is not that interesting, so we can specify a minimizing function so that the result is similar to another dataset. `metamerise` will start from the last metamer of the previous run if the `data` argument is a list of metamers and append the result.
```{r}
x_shape <- subset(datasauRus::datasaurus_dozen, dataset == "x_shape")
x_shape$dataset <- NULL
```
```{r}
metamers <- metamerise(dino,
preserve = mean_cor,
minimize = mean_dist_to(x_shape),
stop_if = minimize_ratio(0.02),
keep = 99)
```
Now the result is a bit more impressive.
```{r}
ggplot(tail(metamers), aes(x, y)) +
geom_point(data = dino, color = "red", alpha = 0.5, size = 0.4) +
geom_point()
```
We can animate the whole thing.
```{r, gganimate = list(fps = 30)}
library(gganimate)
ggplot(metamers, aes(x, y)) +
geom_point() +
transition_manual(.metamer)
```
You can freehand your own starting or target data with the `draw_data()` utility, that will open a shiny interface. You might need to install `shiny` and `miniUI` with `install.packages(c("shiny", "miniUI"))`.
Metamerizing operations can be chained while changing the minimizing function.
```{r}
star <- subset(datasauRus::datasaurus_dozen, dataset == "star")
star$dataset <- NULL
set.seed(42)
metamers <- metamerise(dino,
preserve = mean_cor,
minimize = mean_dist_to(x_shape),
stop_if = minimize_ratio(0.05),
keep = 29) |>
set_minimize(mean_dist_to(star)) |>
metamerise(stop_if = minimize_ratio(0.05),
keep = 30) |>
set_minimize(mean_dist_to(dino)) |>
metamerise(stop_if = minimize_ratio(0.05),
keep = 30)
```
And the full sequence
```{r, gganimate = list(nframes = 30*3, fps = 30)}
ggplot(metamers, aes(x, y)) +
geom_point() +
transition_manual(.metamer)
```