In syunhong/slice: Manipulating Activity Space Data for the Measurement of Segregation
knitr::opts_chunk$set(
collapse = TRUE,
comment = "#>"
)
Introduction
The activity space is the concept of a social place that individuals visit on
a daily basis, and the concept of an activity space itself can be thought of a
lot on a daily basis, but it has not been dealt with largely in the study.
However, many studies have recently been conducted in that the analysis based
on the activity space can realistically observe the segregation and disconnection
experienced by social members in real life compared to the existing
residential-centered approach(Wong and Shaw, 2011; Farber et al., 2012;
ang et al., 2012; Wang and Li, 2016). However, this activity-space-based
research has several problems to be overcome. In order to deal with the activity
space, it is necessary to deal with data in a more complex form than the
existing data, which causes difficulties in analysis and takes a lot of time.
The slice package was developed to make it easier to handle activity space data
in R in this flow. The slice package can find out what area people are located
at a specific time based on people's activity space, based on each travel record,
and has a function that makes it easier to handle such travel records.
In this Vignette, we will use the 2016 Korea Transport Database of South
Korea to deal with how and where people are distributed in each time zone
through slice package.
library(devtools)
load_all()
Classes
Before actually starting, I will explain the two classes included in the package
with data. The slice package has two classes, the names of these two classes are
ASpace and ASpaces, respectively. First, let's proceed with the explanation
while looking at the data.
data("Krtrdata")
class(korea2016)
##ASpaces class
korea2016@data[[1]]
korea2016@attr
##Not run
#korea2016@sp
First is the ASpaces class. The ASpaces class has a total of three slots,
each of which is 'data', 'attr', and 'sp'.
-
The slot 'data' contains activity space information for us to analyze,
and each one is entered into the list inside the data slot.
-
The slot 'attr' is the metadata of the ASpaces class, and contains the
time when the data was collected or the name of the data.
-
Finally, the slot 'sp' contains SpatialPolygon or SpatialPoint that contains the
spatial extent of the activity space.
Note that the space unit or spatial range of the data inside the sp
slot must match the data inside the data slot.
class(korea2016@data[[1]])
##ASpace class
korea2016@data[[1]]@info
korea2016@data[[1]]@trip
Next is the ASpace class. Unlike the ASpaces class, ASpace has one active space
and demographic data, and has a total of two slots. Each slot is Info and Trip.
-
The slot 'info' contains demographic and socioeconomic variables
of the respondent as a list object.
-
The slot 'trip' contains information on trips made by the survey respondent.
In other words, if you briefly describe these two classes,
you can say that the ASpace class is a class that has data for each person,
and ASpaces is a class that has multiple data as its name suggests.
When looking at the data inside ASpace, the data starts with each individual
number and has demographic characteristics such as gender and income level in
the residential area, and inside the trip, it is possible to check the
characteristcs associated with travel.
Note that we have looked at the data, let's see how we can actually
analyze the activity space based on this class.
Measuring spatial distribution
First of all, in the simplest way, it will be possible to measure whether the
degree of distribution of people can vary by time zone. As the name suggests,
the slice function slices consecutive data in time and allows you to see
a section of time.
The slice function can be easily used as follows.
slice(x, at)
In this syntax, enter the ASpaces class for x and enter your desired time for at.
Like this, if you want to check the population distribution at 2 p.m. of the
data you have, the function can be entered as follows.
slice(korea2016, 1400)
Of course, you can simply input this, but it is expected that the operation time
will take a long time due to the size of the data, so let's make it possible to
perform faster operation by entering an additional statement.
It was adjusted to enable parallelized operation through the mc argument, and
the function was controlled to operate in a total of 4 cores through the 'core'
argument. The result appears as follows.
library(parallel)
all1400 <- slice(korea2016, 1400, silent = TRUE, mc = TRUE, core = 4)
head(all1400)
nrow(all1400)
When you check the result, you can see that the format has changed compared to
the previous one, and also that several variables have been added. In other
words, a result in the form of data.frame indicating the purpose of which
individuals with demographic characteristics on the left arrived at the location
by means of transportation for what purpose was derived. And through the last
on.move variable, you can also check whether the person is currently moving or
is active in the area.
However, since the resulting data frame contains too much data (108022),
it is difficult to know which regions are distributed more and which regions
are less distributed. Therefore, we will refine the resulting data and find out
how much it is distributed by region through the slice2df function.
The simple usage of slice2df function is as follows.
slice2df(x, var1)
In this syntax, the data frame that is the result of slice is input in x, and
the location of data.frame is input in var.
head(slice2df(all1400, "location"))
Like this, you can check how many people are active by region at that time.
slice2df also uses an additional variable You can check the distribution of the
population by region in more detail.
result1 <- slice2df(all1400, "location", "sex")
head(slice2df(all1400, "location", "sex"))
Of course, you can check the results more easily than before in this way,
but the table is still too long to check the results at once.
It is true that it is still difficult. From now on, we will visualize with the
data of sp class inside ASpaces.
Visualization
The result of slice can be visualized on the premise that the space unit of sp
inside ASpaces is the same as the unit of activity space.
First, merge the result and sp data for visualization.
korea2016@sp@data <- merge(korea2016@sp@data,
result1, by.x = "adm_cd", by.y = "Area")
head(korea2016@sp@data)
Since the number of population counted by region has been well entered into the
spatial data, a simple visualization can now proceed.
The code for visualization is as follows.
library(RColorBrewer)
library(classInt)
## Specify par so that three maps fit on one screen
par(mar = c(1, 1, 1, 1))
par(mai=c(1, 0, 1, 0))
par(mfrow = c(1, 3))
## Visualization for Male distribution
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 7], 5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Male", border = F)
## Visualization for Female distribution
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 8], 5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Female", border = F)
## Visualization for total population distribution
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 7] + korea2016@sp@data[, 8],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Both", border = F)
When checking the visualization results, it was found that there was no
significant difference in the distribution between men and women at 2:00 pm.
Of course, these results may have a large influence on working hours due to the
specificity of time. If so, what would be the result of different times?
library(parallel)
## Set the time to 8:00 am and 20:00 pm
all800 <- slice(korea2016, 800, silent = TRUE, mc = TRUE, core = 4)
all2000 <- slice(korea2016, 2000, silent = TRUE, mc = TRUE, core = 4)
## refining results
result2 <- slice2df(all800, "location", "sex")
result3 <- slice2df(all2000, "location", "sex")
## merging results to sp slot
korea2016@sp@data <- merge(korea2016@sp@data,
result2, by.x = "adm_cd", by.y = "Area")
korea2016@sp@data <- merge(korea2016@sp@data,
result3, by.x = "adm_cd", by.y = "Area")
head(korea2016@sp@data)
par(mar = c(1, 1, 1, 1))
par(mai=c(1, 0, 1, 0))
par(mfrow = c(1, 3))
## Visualization of 8 AM results
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 9],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Male", border = F)
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 10],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Female", border = F)
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 9] + korea2016@sp@data[, 10],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Both", border = F)
par(mar = c(1, 1, 1, 1))
par(mai=c(1, 0, 1, 0))
par(mfrow = c(1, 3))
## Visualization of 8 PM results
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 11],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Male", border = F)
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 12],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Female", border = F)
my_colors <- brewer.pal(5, "Reds")
class_of_freq <- classIntervals(korea2016@sp@data[, 11] + korea2016@sp@data[, 12],
5, style = "jenks")
colors <- findColours(class_of_freq, my_colors)
plot(korea2016@sp, col = colors, main = "Both", border = F)
In this way, you can see how it changes over time. This can be confirmed as a
result of spatially differentiated residences and jobs in Seoul, Korea. However,
it was confirmed that the difference between women and men still did not appear
noticeably. It can be interpreted that the spatial distribution of men and women
is not very different in terms of activity space.
syunhong/slice documentation built on Feb. 7, 2021, 4:55 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
knitr::opts_chunk$set( collapse = TRUE, comment = "#>" )
Introduction
The activity space is the concept of a social place that individuals visit on a daily basis, and the concept of an activity space itself can be thought of a lot on a daily basis, but it has not been dealt with largely in the study. However, many studies have recently been conducted in that the analysis based on the activity space can realistically observe the segregation and disconnection experienced by social members in real life compared to the existing residential-centered approach(Wong and Shaw, 2011; Farber et al., 2012; ang et al., 2012; Wang and Li, 2016). However, this activity-space-based research has several problems to be overcome. In order to deal with the activity space, it is necessary to deal with data in a more complex form than the existing data, which causes difficulties in analysis and takes a lot of time.
The slice package was developed to make it easier to handle activity space data in R in this flow. The slice package can find out what area people are located at a specific time based on people's activity space, based on each travel record, and has a function that makes it easier to handle such travel records.
In this Vignette, we will use the 2016 Korea Transport Database of South Korea to deal with how and where people are distributed in each time zone through slice package.
library(devtools) load_all()
Classes
Before actually starting, I will explain the two classes included in the package with data. The slice package has two classes, the names of these two classes are ASpace and ASpaces, respectively. First, let's proceed with the explanation while looking at the data.
data("Krtrdata")
class(korea2016) ##ASpaces class korea2016@data[[1]] korea2016@attr ##Not run #korea2016@sp
First is the ASpaces class. The ASpaces class has a total of three slots, each of which is 'data', 'attr', and 'sp'.
-
The slot 'data' contains activity space information for us to analyze, and each one is entered into the list inside the data slot.
-
The slot 'attr' is the metadata of the ASpaces class, and contains the time when the data was collected or the name of the data.
-
Finally, the slot 'sp' contains SpatialPolygon or SpatialPoint that contains the spatial extent of the activity space.
Note that the space unit or spatial range of the data inside the sp slot must match the data inside the data slot.
class(korea2016@data[[1]]) ##ASpace class korea2016@data[[1]]@info korea2016@data[[1]]@trip
Next is the ASpace class. Unlike the ASpaces class, ASpace has one active space and demographic data, and has a total of two slots. Each slot is Info and Trip.
-
The slot 'info' contains demographic and socioeconomic variables of the respondent as a list object.
-
The slot 'trip' contains information on trips made by the survey respondent.
In other words, if you briefly describe these two classes, you can say that the ASpace class is a class that has data for each person, and ASpaces is a class that has multiple data as its name suggests.
When looking at the data inside ASpace, the data starts with each individual number and has demographic characteristics such as gender and income level in the residential area, and inside the trip, it is possible to check the characteristcs associated with travel.
Note that we have looked at the data, let's see how we can actually analyze the activity space based on this class.
Measuring spatial distribution
First of all, in the simplest way, it will be possible to measure whether the degree of distribution of people can vary by time zone. As the name suggests, the slice function slices consecutive data in time and allows you to see a section of time.
The slice function can be easily used as follows.
slice(x, at)
In this syntax, enter the ASpaces class for x and enter your desired time for at. Like this, if you want to check the population distribution at 2 p.m. of the data you have, the function can be entered as follows.
slice(korea2016, 1400)
Of course, you can simply input this, but it is expected that the operation time will take a long time due to the size of the data, so let's make it possible to perform faster operation by entering an additional statement.
It was adjusted to enable parallelized operation through the mc argument, and the function was controlled to operate in a total of 4 cores through the 'core' argument. The result appears as follows.
library(parallel) all1400 <- slice(korea2016, 1400, silent = TRUE, mc = TRUE, core = 4) head(all1400) nrow(all1400)
When you check the result, you can see that the format has changed compared to the previous one, and also that several variables have been added. In other words, a result in the form of data.frame indicating the purpose of which individuals with demographic characteristics on the left arrived at the location by means of transportation for what purpose was derived. And through the last on.move variable, you can also check whether the person is currently moving or is active in the area.
However, since the resulting data frame contains too much data (108022), it is difficult to know which regions are distributed more and which regions are less distributed. Therefore, we will refine the resulting data and find out how much it is distributed by region through the slice2df function.
The simple usage of slice2df function is as follows.
slice2df(x, var1)
In this syntax, the data frame that is the result of slice is input in x, and the location of data.frame is input in var.
head(slice2df(all1400, "location"))
Like this, you can check how many people are active by region at that time. slice2df also uses an additional variable You can check the distribution of the population by region in more detail.
result1 <- slice2df(all1400, "location", "sex") head(slice2df(all1400, "location", "sex"))
Of course, you can check the results more easily than before in this way, but the table is still too long to check the results at once. It is true that it is still difficult. From now on, we will visualize with the data of sp class inside ASpaces.
Visualization
The result of slice can be visualized on the premise that the space unit of sp inside ASpaces is the same as the unit of activity space. First, merge the result and sp data for visualization.
korea2016@sp@data <- merge(korea2016@sp@data, result1, by.x = "adm_cd", by.y = "Area") head(korea2016@sp@data)
Since the number of population counted by region has been well entered into the spatial data, a simple visualization can now proceed.
The code for visualization is as follows.
library(RColorBrewer) library(classInt) ## Specify par so that three maps fit on one screen par(mar = c(1, 1, 1, 1)) par(mai=c(1, 0, 1, 0)) par(mfrow = c(1, 3)) ## Visualization for Male distribution my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 7], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Male", border = F) ## Visualization for Female distribution my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 8], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Female", border = F) ## Visualization for total population distribution my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 7] + korea2016@sp@data[, 8], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Both", border = F)
When checking the visualization results, it was found that there was no significant difference in the distribution between men and women at 2:00 pm. Of course, these results may have a large influence on working hours due to the specificity of time. If so, what would be the result of different times?
library(parallel) ## Set the time to 8:00 am and 20:00 pm all800 <- slice(korea2016, 800, silent = TRUE, mc = TRUE, core = 4) all2000 <- slice(korea2016, 2000, silent = TRUE, mc = TRUE, core = 4) ## refining results result2 <- slice2df(all800, "location", "sex") result3 <- slice2df(all2000, "location", "sex") ## merging results to sp slot korea2016@sp@data <- merge(korea2016@sp@data, result2, by.x = "adm_cd", by.y = "Area") korea2016@sp@data <- merge(korea2016@sp@data, result3, by.x = "adm_cd", by.y = "Area") head(korea2016@sp@data)
par(mar = c(1, 1, 1, 1)) par(mai=c(1, 0, 1, 0)) par(mfrow = c(1, 3)) ## Visualization of 8 AM results my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 9], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Male", border = F) my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 10], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Female", border = F) my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 9] + korea2016@sp@data[, 10], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Both", border = F)
par(mar = c(1, 1, 1, 1)) par(mai=c(1, 0, 1, 0)) par(mfrow = c(1, 3)) ## Visualization of 8 PM results my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 11], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Male", border = F) my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 12], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Female", border = F) my_colors <- brewer.pal(5, "Reds") class_of_freq <- classIntervals(korea2016@sp@data[, 11] + korea2016@sp@data[, 12], 5, style = "jenks") colors <- findColours(class_of_freq, my_colors) plot(korea2016@sp, col = colors, main = "Both", border = F)
In this way, you can see how it changes over time. This can be confirmed as a result of spatially differentiated residences and jobs in Seoul, Korea. However, it was confirmed that the difference between women and men still did not appear noticeably. It can be interpreted that the spatial distribution of men and women is not very different in terms of activity space.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.