| Title: | Create Data with Identical Statistics |
| Version: | 0.3.0 |
| Description: | Creates data with identical statistics (metamers) using an iterative algorithm proposed by Matejka & Fitzmaurice (2017) <doi:10.1145/3025453.3025912>. |
| URL: | https://eliocamp.github.io/metamer/ |
| BugReports: | https://github.com/eliocamp/metamer/issues |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| ByteCompile: | yes |
| Language: | en-US |
| Depends: | R (≥ 2.10) |
| Imports: | FNN, progress (≥ 1.2.0), methods |
| Suggests: | shiny, miniUI, testthat (≥ 2.1.0), data.table, covr, sf |
| RoxygenNote: | 7.2.0 |
| NeedsCompilation: | no |
| Packaged: | 2022-06-23 19:55:43 UTC; elio |
| Author: | Elio Campitelli 
[cre, aut] |
| Maintainer: | Elio Campitelli <elio.campitelli@cima.fcen.uba.ar> |
| Repository: | CRAN |
| Date/Publication: | 2022-06-23 20:10:01 UTC |
metamer: Create Data with Identical Statistics
Description
Creates data with identical statistics (metamers) using an iterative algorithm proposed by Matejka & Fitzmaurice (2017) doi:10.1145/3025453.3025912.
Overview
Create metamers with the metamerize() function.
Some helper functions included:
-
draw_data()for drawing 2D datasets by hand anddensify()for increasing the point density of those drawings. -
delayed_with()for defining statistics to preserve. -
moments_n()for preserving moments of order n. -
mean_dist_to()for minimizing the mean distance to a known target dataset.
The as.data.frame()/[data.table::as.data.table()] methods included will turn a metamer_list' into a tidy
data.frame.
Inspired by Matejka & Fitzmaurice (2017) awesome paper.
Author(s)
Maintainer: Elio Campitelli elio.campitelli@cima.fcen.uba.ar (ORCID)
References
Matejka, J., & Fitzmaurice, G. (2017). Same Stats, Different Graphs. Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems - CHI ’17, 1290–1294. https://doi.org/10.1145/3025453.3025912
See Also
Useful links:
Set metamer parameters
Description
Set metamer parameters
Usage
clear_minimize(metamer_list)
clear_minimise(metamer_list)
set_minimise(metamer_list, minimize)
set_minimize(metamer_list, minimize)
get_last_metamer(metamer_list)
set_annealing(metamer_list, annealing)
set_perturbation(metamer_list, perturbation)
set_perturbation(metamer_list, perturbation)
set_start_probability(metamer_list, start_probability)
set_K(metamer_list, K)
set_change(metamer_list, change)
Arguments
metamer_list |
A |
minimize |
An optional function to minimize in the process. Must take the data as argument and return a single numeric. |
annealing |
Logical indicating whether to perform annealing. |
perturbation |
Numeric with the magnitude of the random perturbations.
Can be of length 1 or |
start_probability |
initial probability of rejecting bad solutions. |
K |
speed/quality tradeoff parameter. |
change |
A character vector with the names of the columns that need to be changed. |
Apply expressions to data.frames
Description
Creates a function that evaluates expressions in a future data.frame. Is like
with(), but the data argument is passed at a later step.
Usage
delayed_with(...)
Arguments
... |
Expressions that will be evaluated. |
Details
Each expression in ... must return a single numeric value. They can be named or
return named vectors.
Value
A function that takes a data.frame and returns the expressions in ...
evaluated in an environment constructed from it.
See Also
Other helper functions:
densify(),
draw_data(),
mean_dist_to_sf(),
mean_dist_to(),
mean_self_proximity(),
moments_n(),
truncate_to()
Examples
some_stats <- delayed_with(mean_x = mean(x), mean(y), sd(x), coef(lm(x ~ y)))
data <- data.frame(x = rnorm(20) , y = rnorm(20))
some_stats(data)
Increase resolution of data
Description
Interpolates between the output of draw_data() and increases the point
density of each stroke.Useful for avoiding sparse targets that result in
clumping of points when metamerizing. It only has an effect on strokes (made
by double clicking).
Usage
densify(data, res = 2)
Arguments
data |
A |
res |
A numeric indicating the multiplicative resolution (i.e. 2 = double resolution). |
Value
A data.frame with the x and y values of your data and a .group column
that identifies each stroke.
See Also
Other helper functions:
delayed_with(),
draw_data(),
mean_dist_to_sf(),
mean_dist_to(),
mean_self_proximity(),
moments_n(),
truncate_to()
Freehand drawing
Description
Opens up a dialogue that lets you draw your data.
Usage
draw_data(data = NULL)
Arguments
data |
Optional |
Value
A data.frame with the x and y values of your data and a .group column
that identifies each stroke.
See Also
Other helper functions:
delayed_with(),
densify(),
mean_dist_to_sf(),
mean_dist_to(),
mean_self_proximity(),
moments_n(),
truncate_to()
Mean minimum distance
Description
Creates a function to get the mean minimum distance between two sets of points.
Usage
mean_dist_to(target, squared = TRUE)
Arguments
target |
A |
squared |
Logical indicating whether to compute the mean squared
distance (if |
Value
A function that takes a data.frame with the same number of columns as
target and then returns the mean minimum distance between them.
See Also
Other helper functions:
delayed_with(),
densify(),
draw_data(),
mean_dist_to_sf(),
mean_self_proximity(),
moments_n(),
truncate_to()
Examples
target <- data.frame(x = rnorm(100), y = rnorm(100))
data <- data.frame(x = rnorm(100), y = rnorm(100))
distance <- mean_dist_to(target)
distance(data)
Mean distance to an sf object
Description
Mean distance to an sf object
Usage
mean_dist_to_sf(target, coords = c("x", "y"), buffer = 0, squared = TRUE)
Arguments
target |
An sf object. |
coords |
Character vector with the columns of the data object that define de coordinates. |
buffer |
Buffer around the sf object. Distances smaller
than |
squared |
Logical indicating whether to compute the mean squared
distance (if |
See Also
Other helper functions:
delayed_with(),
densify(),
draw_data(),
mean_dist_to(),
mean_self_proximity(),
moments_n(),
truncate_to()
Inverse of the mean self distance
Description
Returns the inverse of the mean minimum distance between different pairs of points. It's intended to be used as a minimizing function to, then, maximize the distance between points.
Usage
mean_self_proximity(data)
Arguments
data |
a data.frame |
See Also
Other helper functions:
delayed_with(),
densify(),
draw_data(),
mean_dist_to_sf(),
mean_dist_to(),
moments_n(),
truncate_to()
Create metamers
Description
Produces very dissimilar datasets with the same statistical properties.
Usage
metamerise(
data,
preserve,
minimize = NULL,
change = colnames(data),
round = truncate_to(2),
stop_if = n_tries(100),
keep = NULL,
annealing = TRUE,
K = 0.02,
start_probability = 0.5,
perturbation = 0.08,
name = "",
verbose = interactive()
)
metamerize(
data,
preserve,
minimize = NULL,
change = colnames(data),
round = truncate_to(2),
stop_if = n_tries(100),
keep = NULL,
annealing = TRUE,
K = 0.02,
start_probability = 0.5,
perturbation = 0.08,
name = "",
verbose = interactive()
)
new_metamer(data, preserve, round = truncate_to(2))
Arguments
data |
A |
preserve |
A function whose result must be kept exactly the same. Must take the data as argument and return a numeric vector. |
minimize |
An optional function to minimize in the process. Must take the data as argument and return a single numeric. |
change |
A character vector with the names of the columns that need to be changed. |
round |
A function to apply to the result of |
stop_if |
A stopping criterium. See n_tries. |
keep |
Max number of metamers to return. |
annealing |
Logical indicating whether to perform annealing. |
K |
speed/quality tradeoff parameter. |
start_probability |
initial probability of rejecting bad solutions. |
perturbation |
Numeric with the magnitude of the random perturbations.
Can be of length 1 or |
name |
Character for naming the metamers. |
verbose |
Logical indicating whether to show a progress bar. |
Details
It follows Matejka & Fitzmaurice (2017) method of constructing metamers.
Beginning from a starting dataset, it iteratively adds a small perturbation,
checks if preserve returns the same value (up to signif significant digits)
and if minimize has been lowered, and accepts the solution for the next
round. If annealing is TRUE, it also accepts solutions with bigger
minimize with an ever decreasing probability to help the algorithm avoid
local minimums.
The annealing scheme is adapted from de Vicente et al. (2003).
If data is a metamer_list, the function will start the algorithm from the
last metamer of the list. Furthermore, if preserve and/or minimize
are missing, the previous functions will be carried over from the previous call.
minimize can be also a vector of functions. In that case, the process minimizes
the product of the functions applied to the data.
Value
A metamer_list object (a list of data.frames).
References
Matejka, J., & Fitzmaurice, G. (2017). Same Stats, Different Graphs. Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems - CHI ’17, 1290–1294. https://doi.org/10.1145/3025453.3025912 de Vicente, Juan, Juan Lanchares, and Román Hermida. (2003). ‘Placement by Thermodynamic Simulated Annealing’. Physics Letters A 317(5): 415–23.
See Also
delayed_with() for a convenient way of making functions suitable for
preserve, mean_dist_to() for a convenient way of minimizing the distance
to a known target in minimize, mean_self_proximity() for maximizing the
"self distance" to prevent data clumping.
Examples
data(cars)
# Metamers of `cars` with the same mean speed and dist, and correlation
# between the two.
means_and_cor <- delayed_with(mean_speed = mean(speed),
mean_dist = mean(dist),
cor = cor(speed, dist))
set.seed(42) # for reproducibility.
metamers <- metamerize(cars,
preserve = means_and_cor,
round = truncate_to(2),
stop_if = n_tries(1000))
print(metamers)
last <- tail(metamers)
# Confirm that the statistics are the same
cbind(original = means_and_cor(cars),
metamer = means_and_cor(last))
# Visualize
plot(tail(metamers))
points(cars, col = "red")
Compute moments
Description
Returns a function that will return uncentered moments
Usage
moments_n(orders, cols = NULL)
Arguments
orders |
Numeric with the order of the uncentered moments that will be computed. |
cols |
Character vector with the name of the columns of the data for which
moments will be computed. If |
Value
A function that takes a data.frame and return a named numeric vector of the
uncentered moments of the columns.
See Also
Other helper functions:
delayed_with(),
densify(),
draw_data(),
mean_dist_to_sf(),
mean_dist_to(),
mean_self_proximity(),
truncate_to()
Examples
data <- data.frame(x = rnorm(100), y = rnorm(100))
moments_3 <- moments_n(1:3)
moments_3(data)
moments_3 <- moments_n(1:3, "x")
moments_3(data)
Stop conditions
Description
Stop conditions
Usage
n_tries(n)
n_metamers(n)
minimize_ratio(r)
Arguments
n |
integer number of tries or metamers. |
r |
Ratio of minimize value to shoot for. If |
Rounding functions
Description
Rounding functions
Usage
truncate_to(digits)
round_to(digits)
Arguments
digits |
Number of significant digits. |
See Also
Other helper functions:
delayed_with(),
densify(),
draw_data(),
mean_dist_to_sf(),
mean_dist_to(),
mean_self_proximity(),
moments_n()
Other helper functions:
delayed_with(),
densify(),
draw_data(),
mean_dist_to_sf(),
mean_dist_to(),
mean_self_proximity(),
moments_n()