Alluvial diagrams consist of multiple horizontally-distributed columns (axes) representing factor variables, vertical divisions (strata) of these axes representing these variables' values; and splines (alluvial flows) connecting vertical subdivisions (lodes) within strata of adjacent axes representing subsets or amounts of observations that take the corresponding values of the corresponding variables. This function checks a data frame for either of two types of alluvial structure:
is_lodes_form(data, key, value, id, weight = NULL, logical = TRUE,
silent = FALSE)is_alluvia_form(data, ..., axes = NULL, weight = NULL,
logical = TRUE, silent = FALSE)
to_lodes_form(data, ..., axes = NULL, key = "x", value = "stratum",
id = "alluvium", diffuse = FALSE, discern = FALSE)
to_alluvia_form(data, key, value, id, distill = FALSE)
A data frame.
In to_lodes_form, handled as in
tidyr::gather() and used to name the new axis (key), stratum
(value), and alluvium (identifying) variables. In to_alluvia_form,
handled as in tidyr::spread() and used to identify the fields
of data to be used as the axis (key), stratum (value), and alluvium
(identifying) variables.
Optional field of data, handled using
rlang::enquo(), to be used as heights or depths of the alluvia
or lodes.
Deprecated. Whether to return a logical value or a character string indicating the type of alluvial structure ("none", "lodes", or "alluvia").
Whether to print messages.
Used in is_alluvia_form and to_lodes_form as in
dplyr::select() to determine axis variables, as an alternative
to axes. Ignored when axes is provided.
In *_alluvia_form, handled as in
dplyr::select() and used to identify the field(s) of
data to be used as axes.
Fields of data, handled using
tidyselect::vars_select(), to merge into the reshapen data by
id. They must be a subset of the axis variables. Alternatively, a
logical value indicating whether to merge all (TRUE) or none
(FALSE) of the axis variables.
Logical value indicating whether to suffix values of the
variables used as axes that appear at more than one variable in order to
distinguish their factor levels. This forces the levels of the combined
factor variable value to be in the order of the axes.
A logical value indicating whether to include variables, other
than those passed to key and value, that vary within values
of id. Alternatively, a function (or its name) to be used to distill
each such variable to a single value. In addition to existing functions,
distill accepts the character values "first" (used if
distill is TRUE), "last", and "most" (which
returns the modal value).
One row per lode, wherein each row encodes a subset or amount of
observations having a specific profile of axis values, a key field
encodes the axis, a value field encodes the value within each axis, and a
id column identifies multiple lodes corresponding to the same subset or
amount of observations. is_lodes_form tests for this structure.
One row per alluvium, wherein each row encodes a subset or amount of
observations having a specific profile of axis values and a set axes of
fields encodes its values at each axis variable. is_alluvia_form tests
for this structure.
to_lodes_form takes a data frame with several designated variables to
be used as axes in an alluvial diagram, and reshapes the data frame so that
the axis variable names constitute a new factor variable and their values
comprise another. Other variables' values will be repeated, and a
row-grouping variable can be introduced. This function invokes
tidyr::gather().
to_alluvia_form takes a data frame with axis and axis value variables
to be used in an alluvial diagram, and reshape the data frame so that the
axes constitute separate variables whose values are given by the value
variable. This function invokes tidyr::spread().
Other alluvial data manipulation: self-adjoin
# NOT RUN {
# Titanic data in alluvia format
titanic_alluvia <- as.data.frame(Titanic)
head(titanic_alluvia)
is_alluvia_form(titanic_alluvia,
weight = "Freq")
# Titanic data in lodes format
titanic_lodes <- to_lodes_form(titanic_alluvia,
key = "x", value = "stratum", id = "alluvium",
axes = 1:4)
head(titanic_lodes)
is_lodes_form(titanic_lodes,
key = "x", value = "stratum", id = "alluvium",
weight = "Freq")
# again in lodes format, this time diffusing the `Class` variable
titanic_lodes2 <- to_lodes_form(titanic_alluvia,
key = variable, value = value,
id = passenger,
1:3, diffuse = Class)
head(titanic_lodes2)
is_lodes_form(titanic_lodes2,
key = variable, value = value, id = passenger,
weight = Freq)
# curriculum data in lodes format
data(majors)
head(majors)
is_lodes_form(majors,
key = "semester", value = "curriculum", id = "student")
# curriculum data in alluvia format
majors_alluvia <- to_alluvia_form(majors,
key = "semester", value = "curriculum",
id = "student")
head(majors_alluvia)
is_alluvia_form(majors_alluvia, tidyselect::starts_with("CURR"))
# distill variables that vary within `id` values
set.seed(1)
majors$hypo_grade <- LETTERS[sample(5, size = nrow(majors), replace = TRUE)]
majors_alluvia2 <- to_alluvia_form(majors,
key = "semester", value = "curriculum",
id = "student",
distill = "most")
head(majors_alluvia2)
# options to distinguish strata at different axes
gg <- ggplot(majors_alluvia,
aes(axis1 = CURR1, axis2 = CURR7, axis3 = CURR13))
gg +
geom_alluvium(aes(fill = as.factor(student)), width = 2/5, discern = TRUE) +
geom_stratum(width = 2/5, discern = TRUE) +
geom_text(stat = "stratum", discern = TRUE, label.strata = TRUE)
gg +
geom_alluvium(aes(fill = as.factor(student)), width = 2/5, discern = FALSE) +
geom_stratum(width = 2/5, discern = FALSE) +
geom_text(stat = "stratum", discern = FALSE, label.strata = TRUE)
# warning when inappropriate
ggplot(majors[majors$semester %in% paste0("CURR", c(1, 7, 13)), ],
aes(x = semester, stratum = curriculum, alluvium = student,
label = curriculum)) +
geom_alluvium(aes(fill = as.factor(student)), width = 2/5, discern = TRUE) +
geom_stratum(width = 2/5, discern = TRUE) +
geom_text(stat = "stratum", discern = TRUE)
# }
Run the code above in your browser using DataLab