R/missing_values.R
In antonmalko/ettools: Utility functions for dealing with eye-tracking data
Defines functions
check_columns_existence
check_rois_mois_existence
zero_inflate_cells_count
count_cells
count_NAs
count_extremes
plot_cells_heatmap
plot_cells_count
report_NAs_count
report_extremes_count
Documented in
check_columns_existence
check_rois_mois_existence
count_cells
count_extremes
count_NAs
plot_cells_count
plot_cells_heatmap
report_extremes_count
report_NAs_count
zero_inflate_cells_count
# ---- Helper functions ----
check_columns_existence <- function(dat, columns){
#' Check whether specified columns exist in a data frame
#'
#' @param dat data.frame to check for the existence of columns
#' @param columns a vector of column names (should be "character")
#' @return Throws an error if some of the columns are not present in the
#' data.frame. The error text indicates which columns are missing
if (!all(columns %in% colnames(dat))){
stop(paste("Some columns are not found in the data: ",
paste(columns[!columns %in% colnames(dat)])))
}
}
check_rois_mois_existence <- function(dat, rois, mois, region.col, measure.col){
#' Check whether eye-tracking data contains regions of interest (rois) and
#' measures of interest (mois).
#'
#' @param dat data.frame to check the existence of rois and mois
#' @param rois,mois vectors of region and measures identifiers, which will be
#' checked against \code{region.col} and \code{measure.col} respectively.
#' The type of entries in the vector should be the same as in the corresponding
#' columns
#' @param region.col,measure.col name of column containing region
#' and eye-tracking measures identifiers respectively. See "Details" for more
#' info.
#'
#' @details
#' The type of \code{region.col} and \code{measure.col} should be either
#' "character" or "quosure" (see documentation for "quosure" in \code{rlang}
#' package)
#' @return Throws an error if some of the rois and mois are not present in the
#' data.frame; the error text indicates which rois/mois are missing.
#'
regions.in.the.data <- dat[[rlang::quo_name(region.col)]]
measures.in.the.data <- dat[[rlang::quo_name(measure.col)]]
if (any(!rois %in% regions.in.the.data)){
stop(paste0("Some rois are not found in the data: ",
paste(rois[!rois %in% regions.in.the.data], collapse = ", ")))
}
if (any(!mois %in% measures.in.the.data)){
stop(paste0("Some measures of interest are not found in the data: ",
paste(mois[!mois %in% measures.in.the.data], collapse = ",")))
}
}
zero_inflate_cells_count <- function(dat, by,
cells.count, rois, mois){
#' Make sure that every roi + moi combination has counts
#'
#' This function ensures that every combination of region and measure
#' contains counts. Basically, it takes care fo the situation when a given
#' region-measure doesn't contain cells of the required kind (e.g. NAs or
#' extreme observations). In such cases this function adds 0 as count to make
#' sure that all region-measure combination are present in the summary plot.
#'
#' @param dat data.frame containing information about \code{by} column
#' @param by quosure indicating which column we are aggregating along. Usually
#' it is the subject or the item column.
#' @param cells.count data.frame with cells count data produced by \code{count_cells}.
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted
#'
#' @return data.frame containing the original cells counts as well as added
#' 0 counts
#' @importFrom dplyr %>%
by <- rlang::quo_name(by)
# if there are no missing cells for some combination of region and measure,
# insert 0s; otherwise that column will be completely missing
# expand.grid varies the first column fastest - this is what we need, but the
# order is wrong: e.g. we want spillover_ff, not, ff_spillover...
reg.meas.combs <- expand.grid(mois, rois)
# so we exchange columns
reg.meas.combs <- reg.meas.combs[c(2,1)]
# then paste them together; mdply adds a new column with the result,
# and we only need that, so we only take the third column
reg.meas.combs <- plyr::mdply(reg.meas.combs, paste, sep="_")[[3]]
# create a dummy data.frame with all necessary region+measure combinations
# filled with 0s
dummy.df <- data.frame(tmp = unique(dat[[by]]),
cell = rep(reg.meas.combs,
each = length(unique(dat[[by]]))),
count = 0,
stringsAsFactors = FALSE)
colnames(dummy.df)[1] <- by
# convert to character, otherwise bind_rows_ will throw a warning.
# It's harmless, but we'd better avoi it not to worry the end user unduly
cells.count$cell <- as.character(cells.count$cell)
cells.count <- dplyr::bind_rows(dummy.df, cells.count) %>%
dplyr::group_by_(by, "cell") %>%
dplyr::summarize(count = sum(count)) %>%
dplyr::ungroup()
cells.count$cell <- factor(cells.count$cell,
levels = reg.meas.combs)
return(cells.count)
}
# ----- Counting functions ----
count_cells <- function(dat, by, ..., cast.formula,
value.name = "count") {
#' Count number of observations of specific kind
#'
#' A generic function to count number of certain observations (e.g. NAs or
#' extreme values) by cells (e.g. combination of region and eye-tracking measure).
#'
#' @param dat data.frame containing data to process
#' @param by unquoted column name for the main grouping variable. See "Details"
#' @param ... unqouted column names - other grouping variables. See "Details".
#' @param cast.formula character. A formula for \code{dcast()} (from 'reshape2' package).
#' LHS is the grouping variable, RHS defines cells
#' @param value.name character. name of the column containing counts in the
#' result data.frame
#'
#' @details
#' The function assumes that \code{dat} argument contains the data which has
#' already been subsetted to only contain values of interest. E.g., if you
#' want to count NAs, before passing the data set to this function, you need
#' to filter all non-NAs out. The package contains two specific convenience
#' functions which would do the subseetting for you (the names are self-explanatory):
#' \code{count_NAs} and \code{count_extremes}.
#'
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which will be displayed on the y axis
#' in the summary plots.
#'
#' The column names in \code{...} argument define which columns will act as
#' grouping variables for \code{dplyr::group_by}, thus defining the smallest
#' subset of the data in which the observations should be counted. In a typical
#' use for eye-tracking data, there will be two components here: region and
#' measure columns names. In general, the values in the first item will be
#' varied slowest, the values in the last item will be varied fastest.
#'
#' Thus, e.g. (if \code{by = subj} and \code{... = region.col, measure,col}),
#' the observations will be counted in each region, in each measure,
#' for each subject.
#'
#'@return
#'
#' data.frame with 3 columns (default names):
#'
#' 1. Name of the \code{by} argument, typically subj or item; contains
#' the unique values from the corresponding column in \code{dat}
#' 2. \code{cell}; contains data subset identifiers. In a typical use,
#' if used with regions and measures, it would be region_measure, e.g. "critical_tt"
#' 3. \code{count} count of observations in each subset of the data.
#'
#'@seealso \code{\link{count_NAs}}, \code{\link{count_extremes}}
#'
#' @importFrom dplyr %>%
#' @importFrom rlang !!
#' @importFrom rlang !!!
grouping_vars <- rlang::list2(...)
id.var <- rlang::quo_name(by)
if (!is.data.frame(dat)){
stop("`dat` argument must be a data.frame")
}
if (!plyr::is.formula(cast.formula)){
stop("`cast.formula` argument must be a formula")
}
if (!is.character(value.name)){
stop("`value.name` should be a character")
}
if (!is.character(id.var)){
stop("`id.var` should be a character")
}
if (nrow(dat) == 0){
warning("Cannot count cells: the data is empty")
res <- data.frame(a = numeric(),
b = character(),
c = numeric())
colnames(res) <- c(id.var, "cell", value.name)
return(res)
}
res <- dat %>%
dplyr::group_by(!!!grouping_vars, !!by) %>%
dplyr::summarize(res =n()) %>%
reshape2::dcast(cast.formula, value.var = "res", fill = 0) %>%
reshape2::melt(id.vars = id.var, variable.name = "cell", value.name = value.name)
return(res)
}
count_NAs <- function(dat, by,
rois, mois,
value.col, region.col, measure.col){
#' Convenience wrapper for count_cells() for counting NAs.
#'
#' Count the number of NAs in the data briken down by region, measure and
#' some third factor, usually subject or item.
#'
#' @param dat data.frame containing data to process
#' @param by quosure. column name for the main grouping variable. See "Details"
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted. Any of these can be left unsepcified, in this
#' case, all regions and/or measure form the data will be used.
#' @param value.col quosure. column name containing the NAs (potentially along
#' with non-NA values. Typically it will be the column containing the reaction
#' times or something alike).
#' @param region.col quosure. column name containing region identifiers
#' @param measure.col quosure. column name containing measure identifiers
#'
#' @details
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which varies slowest and will be
#' displayed on the y axis in the summary plots.
#'
#' @return data.frame with 3 columns analogous to \code{\link{count_cells}}
#' (i.e. containing the main grouping variable identifier (typically subject or item),
#' the combination of region and measure identifying subsets of data, and NAs counts)
#' @export
#' @importFrom dplyr %>%
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
rois <- unique(dat[[rlang::quo_name(measure.col)]])
}
check_columns_existence(dat = dat,
columns = c(rlang::quo_name(by),
rlang::quo_name(value.col),
rlang::quo_name(region.col),
rlang::quo_name(measure.col)))
check_rois_mois_existence (dat = dat, rois = rois, mois = mois,
region.col = region.col,
measure.col = measure.col)
filtered.dat <- dat %>%
dplyr::filter(!stats::complete.cases(rlang::UQ(value.col)))
if (nrow(filtered.dat) == 0){
warning(paste0("There are no NAs in column `", rlang::quo_name(value.col), "`"))
res <- data.frame(a = numeric(),
cell = character(),
count = numeric())
colnames(res)[1] <- rlang::quo_name(by)
} else {
filtered.dat <- filtered.dat%>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois)
res <- count_cells(filtered.dat, by = by,
region.col,
measure.col,
cast.formula = stats::formula(paste(rlang::quo_name(by),"~" ,rlang::quo_name(region.col),
" + ", rlang::quo_name(measure.col))))
}
res <- zero_inflate_cells_count(dat = dat, by = by,
cells.count = res,
rois = rois, mois = mois)
return(res)
}
count_extremes <- function(dat, by,
rois, mois,
value.col, region.col, measure.col,
max.cutoff, min.cutoff = 0){
#' Convenience wrapper for \code{\link{count_cells}} for counting NAs.
#'
#' Count the number of NAs in the data briken down by region, measure and
#' some third factor, usually subject or item.
#'
#' @param dat data.frame containing data to process
#' @param by quosure with column name for the main grouping variable. See "Details"
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted. Any of these can be left unsepcified, in this
#' case, all regions and/or measure form the data will be used.
#' @param value.col quosure with column name containing the NAs (potentially along
#' with non-NA values. Typically it will be the column containing the reaction
#' times or something alike).
#' @param region.col quosure with column name containing region identifiers
#' @param measure.col quosure with column name containing measure identifiers
#' @param max.cutoff numeric value, indicating the upper threshold; any values
#' above this threshold will be considered "extreme" and counted
#' @param min.cutoff numeric value, indicating the lower threshold. Defaults to 0
#'
#' @details
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which varies slowest and will be
#' displayed on the y axis in the summary plots.
#'
#' @return data.frame with 4 columns. The first 3 are analogous to
#' \code{\link{count_cells}}
#' (i.e. they contain the main grouping variable identifier (typically subject or item),
#' the combination of region and measure identifying subsets of data,
#' and extreme values counts), and the fourth one, called "direction", indicates
#' whether the value is extremely high (then the corresponding row will contain
#' "above.max" in the"direction" column) or extremely low ("below.min" in the
#' "direction" column).
#' @export
#' @importFrom dplyr %>%
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
rois <- unique(dat[[rlang::quo_name(measure.col)]])
}
check_columns_existence(dat = dat,
columns = c(rlang::quo_name(by),
rlang::quo_name(value.col),
rlang::quo_name(region.col),
rlang::quo_name(measure.col)))
check_rois_mois_existence (dat = dat, rois = rois, mois = mois,
region.col = region.col,
measure.col = measure.col)
dat <- dat %>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois)
high.extremes <- dat %>%
dplyr::filter(rlang::UQ(value.col) > max.cutoff)
low.extremes <- dat %>%
dplyr::filter(rlang::UQ(value.col) < min.cutoff)
if (nrow(high.extremes) == 0){
warning(paste0("There are no observations higher than ", max.cutoff))
high.extremes <- data.frame(a = numeric(),
cell = character(),
count = numeric(),
direction = character())
colnames(high.extremes)[1] <- rlang::quo_name(by)
} else {
high.extremes <- count_cells(high.extremes,
by,
region.col,
measure.col,
cast.formula = stats::formula(paste(rlang::quo_name(by),"~",
rlang::quo_name(region.col), " + ",
rlang::quo_name(measure.col))))
}
if (nrow(low.extremes) == 0){
warning(paste0("There are no observations lower than ", min.cutoff))
low.extremes <- data.frame(a = numeric(),
cell = character(),
count = numeric(), direction = character())
colnames(low.extremes)[1] <- rlang::quo_name(by)
} else {
low.extremes <- count_cells(low.extremes, by,
region.col,
measure.col,
cast.formula = stats::formula(paste(rlang::quo_name(by),"~",
rlang::quo_name(region.col), " + ",
rlang::quo_name(measure.col))))
}
high.extremes <- zero_inflate_cells_count(dat = dat, by = by,
cells.count = high.extremes,
rois = rois, mois = mois)
low.extremes <- zero_inflate_cells_count(dat = dat, by = by,
cells.count = low.extremes,
rois = rois, mois = mois)
high.extremes$direction <- "above.max"
low.extremes$direction <- "below.min"
res <- rbind(high.extremes, low.extremes)
return(res)
}
# ---- Plotting functions ----
plot_cells_heatmap <- function(dat, by, fill, limits = NULL,
plot.name){
#' Plot heatmap of cell counts
#'
#' A generic function to plot heatmaps of cell counts. Shouldn't be used directly
#' to plot counts produced by \code{\link{count_cells}} and related functions, since
#' those counts may lack the data from certain subsets of the data, if those
#' subsets lack the relevant kind of values (e.g. NAs or extreme values).
#' Instead, \code{\link{plot_cells_count}}, or convenience wrappers around it
#' should be called, since they take care of subsets with 0 observations
#' by artificially adding these 0s back to the counts.
#'
#' @param dat data.frame returned by \code{count_NAs} or \code{count_extremes()}.
#' It should have three columns: grouping variable (subj or item),
#' name of the cell (region+fixationtype by default),
#' number of observations in the cell
#' @param by quosure. name of the grouping variable (normally, first column
#' in \code{dat})
#' @param fill quosure. name of the column w/count of observations (normally,
#' third column in \code{dat})
#' @param limits vector of two integers: lowest and highest N of observations
#' possible in the cell to calibrate color coding. If NULL (default),
#' the limits will be set automatically based on the data.
#' @param plot.name character
#' @return heatmap plot of the cell counts
by <- rlang::quo_name(by)
fill <- rlang::quo_name(fill)
check_columns_existence(dat = dat,
columns = c(by,
fill))
# + scale_y_reverse - make the 1st subject/item appear on top;
# + breaks - make sure that each subject has its number shown
out <- ggplot2::ggplot(dat, ggplot2::aes_string(x = "cell", y = by)) +
ggplot2::geom_tile(ggplot2::aes_string(fill = fill), color = "white") +
ggplot2::scale_fill_gradient(low = "white", high = "steelblue",
breaks = seq(min(limits), max(limits), by = 1)) +
ggplot2::geom_text(ggplot2::aes_string(label = fill))+
ggplot2::scale_x_discrete(position = "top") +
ggplot2::ggtitle (plot.name)
if (is.numeric(dat[[by]])){
out <- out + ggplot2::scale_y_reverse(breaks = sort(unique(dat[[by]])))
}
return(out)
}
plot_cells_count <- function(dat, by,
cells.count, rois, mois, fill = count,
plot.name = "Give a sensible name to the plot!"){
#' Plot heatmaps of cell counts
#'
#' A wrapper around \code{\link{plot_cells_heatmap}} which is aware of regions
#' and measures of interest. Because of that it a) adds 0 counts to region + measure
#' combinations which lacked the relevant kind of value (e.g. extreme or NAs);
#' b) visually delimits ROIs in the plot for visual convenience.
#'
#' @param dat data.frame containing information about \code{by} column
#' @param by quosure indicating which column we are aggregating along. Usually
#' it is the subject or the item column.
#' @param cells.count data.frame with cells count data produced by \code{\link{count_cells}}.
#' The function expects that the data.frame has column "cell", containing
#' the names of data subsets for which counts were performed (e.g
#' region+measure combination) and column "count", containing the counts
#' of the relevant values in each cell.
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted
#' @param fill unquoted name of the column with cell counts
#' @param plot.name character
#'
#' @return heatmap plot of the cell counts (expanded compared to the plot
#' returned by \code{\link{plot_cells_heatmap}} - with added zero counts and ROIs
#' delimitation)
#' @export
if (!rlang::is_quosure(by)) {
stop("`by` argument should be a quosure")
}
check_columns_existence(dat = cells.count,
columns = c("cell",
"count"))
fill <- rlang::enquo(fill)
cells.count$cell <- normalize_axes_labels(cells.count$cell)
# determine the max number for fill scale
fill.upper.limit <- max(cells.count$count)
# How many regions and measures? This is used below in geom_vline() statement
# to figure out, how close should the separating vertical lines inserted
n.mois <- length(mois)
n.rois <- length(rois)
cells.plot <- plot_cells_heatmap(cells.count, by = by, fill = fill,
limits = c(0, fill.upper.limit),
plot.name = plot.name)
if (n.rois > 1){
# Insert separating lines between regions for visual convenience. Start after
# first n + 0.5 measures (+0.5 in order to insert the line in between two columns
# of figures), insert them every n, with an upper limit on n =
# n_of_regions * n_of_measures.
cells.plot <- cells.plot +
ggplot2::geom_vline(xintercept = seq(from = n.mois+0.5, to = n.rois*n.mois, by = n.mois))
}
return(cells.plot)
}
# ---- User-level functions ----
report_NAs_count <- function(dat, by = subj,
rois, mois,
value.col = value,
region.col = reg.name, measure.col = fixationtype,
plot.name = "Give a sensible name to the plot!"){
#' Count and plot NAs by regions and measures of interest
#'
#' @param dat data.frame containing the data where the NAs should be counted
#' @param by unquoted column name for the main grouping variable. See "Details"
#' @param rois,mois vectors of region and measures identifiers, for which counts
#' should be produced. The type of entries in the vector should be the same as in
#' \code{region.col} and \code{measure.col} respectively. Any of these can be left
#' unsepcified, in this case, all regions and/or measure form the data will be used.
#' @param value.col unquoted name of the column containing NAs which should
#' be counted (typically it would be the column containing reaction times
#' or something similar)
#' @param region.col unquoted name of the column containing region ids
#' @param measure.col unquoted name of the column containing measure ids
#' @param plot.name character. name for the plot visualizing the counts
#'
#' @details
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which varies slowest and will be
#' displayed on the y axis in the summary plots.
#'
#' @return list with the following components:
#' * counts data.frame containing count data
#' * cells.plot heatmap visualization of the counts
#' @export
#' @importFrom dplyr %>%
value.col <- rlang::enquo(value.col)
region.col <- rlang::enquo(region.col)
measure.col <- rlang::enquo(measure.col)
by <- rlang::enquo(by)
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
mois <- unique(dat[[rlang::quo_name(measure.col)]])
}
cells.count <- dat %>%
dplyr::ungroup() %>%
dplyr::do(., count_NAs(.,
rois = rois, mois = mois,
value.col = value.col, region.col = region.col,
measure.col = measure.col,
by = by))
cells.plot <- plot_cells_count(dat = dat,
cells.count = cells.count,
rois = rois, mois = mois,
by = by, plot.name = plot.name)
return(list(counts = cells.count,
plot = cells.plot))
}
report_extremes_count <- function(dat, by = subj,
rois, mois,
value.col = value,
region.col = reg.name, measure.col = fixationtype,
max.cutoff, min.cutoff = 0,
plot.name = "Give a sensible name to the plot!"){
#' Count and plot extreme observation by regions and measures of interest
#'
#' @param dat data.frame containing information about \code{by} column
#' @param by quosure indicating which column we are aggregating along. Usually
#' it is the subject or the item column.
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted. The type of entries in the vector should be the same as in
#' \code{region.col} and \code{measure.col} respectively.
#' Any of these can be left unsepcified, in this case, all regions and/or measure
#' form the data will be used.
#' @param value.col unquoted name of the column containing NAs which should
#' be counted (typically it would be the column containing reaction times
#' or something similar)
#' @param region.col unquoted name of the column containing region ids
#' @param measure.col unquoted name of the column containing measure ids
#' @param max.cutoff numeric value, indicating the upper threshold; any values
#' above this threshold will be considered "extreme" and counted
#' @param min.cutoff numeric value, indicating the lower threshold. Defaults to 0
#' @param plot.name Name of the heatmap plot with extreme values counts
#'
#' @return list with 3 components:
#' * \code{counts} - data.frame with counts data
#' * \code{count.plots} - list with 3 components
#' * \code{high.plot} - plot with counts of extreme observations above \code{max.cutoff}
#' * \code{low.plot} - plot with counts of extreme observations below \code{min.cutoff}
#' * \code{all.plot} - plot with combined counts of both high and low extreme values
#' * \code{values.plots} - list with 2 components
#' * \code{high.plot} - plot with numerical values of extreme observations above \code{max.cutoff}
#' * \code{low.plot} - plot with numerical values of extreme observations below \code{min.cutoff}
#' @export
#' @importFrom dplyr %>%
value.col <- rlang::enquo(value.col)
region.col <- rlang::enquo(region.col)
measure.col <- rlang::enquo(measure.col)
by <- rlang::enquo(by)
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
mois <- unique(dat[[rlang::quo_name(measure.col)]])
}
# count extreme values
cells.count <- dat %>%
dplyr::do(., count_extremes(., by = by,
rois = rois, mois = mois,
value.col = value.col,
region.col = region.col,
measure.col = measure.col,
max.cutoff = max.cutoff,
min.cutoff = min.cutoff))
count.plots <- list()
values.plots <- list()
# plot values above the max cutoff
count.plots$high <- cells.count %>%
dplyr::filter(direction == "above.max") %>%
plot_cells_count(., dat = dat,
rois = rois, mois = mois,
by = by, plot.name = paste0(plot.name, "_above_", max.cutoff))
# plot values below the min cutoff
count.plots$low <- cells.count %>%
dplyr::filter(direction == "below.min") %>%
plot_cells_count(., dat = dat,
rois = rois, mois = mois,
by = by, plot.name = paste0(plot.name, "_below_", min.cutoff))
# combine the two plots above in one
count.plots$all <- cells.count %>%
dplyr::group_by_(by, "cell") %>%
dplyr::summarize(count = sum(count)) %>%
dplyr::ungroup() %>%
plot_cells_count(., dat = dat,
rois = rois, mois = mois,
by = by, plot.name = paste0(plot.name, "_above_", max.cutoff,
"_and_below_", min.cutoff))
# plot numerical values above the max cutoff
values.plots$high <- dat %>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois) %>%
dplyr::filter(rlang::UQ(value.col) > max.cutoff) %>%
ggplot2::ggplot() +
ggplot2::geom_point(ggplot2::aes_string(y = rlang::quo_name(value.col),
x = rlang::quo_name(measure.col)))+
ggplot2::facet_wrap(stats::formula(paste("~", rlang::quo_name(region.col))))
# plot numerical values below the min cutoff
values.plots$low <- dat %>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois) %>%
dplyr::filter(rlang::UQ(value.col) < min.cutoff) %>%
ggplot2::ggplot() +
ggplot2::geom_point(ggplot2::aes_string(y = rlang::quo_name(value.col),
x = rlang::quo_name(measure.col)))+
ggplot2::facet_wrap(stats::formula(paste("~", rlang::quo_name(region.col))))
return(list(counts = cells.count,
count.plots = count.plots,
values.plots = values.plots))
}
antonmalko/ettools documentation built on May 28, 2019, 3:35 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.
Defines functions check_columns_existence check_rois_mois_existence zero_inflate_cells_count count_cells count_NAs count_extremes plot_cells_heatmap plot_cells_count report_NAs_count report_extremes_count
Documented in check_columns_existence check_rois_mois_existence count_cells count_extremes count_NAs plot_cells_count plot_cells_heatmap report_extremes_count report_NAs_count zero_inflate_cells_count
# ---- Helper functions ----
check_columns_existence <- function(dat, columns){
#' Check whether specified columns exist in a data frame
#'
#' @param dat data.frame to check for the existence of columns
#' @param columns a vector of column names (should be "character")
#' @return Throws an error if some of the columns are not present in the
#' data.frame. The error text indicates which columns are missing
if (!all(columns %in% colnames(dat))){
stop(paste("Some columns are not found in the data: ",
paste(columns[!columns %in% colnames(dat)])))
}
}
check_rois_mois_existence <- function(dat, rois, mois, region.col, measure.col){
#' Check whether eye-tracking data contains regions of interest (rois) and
#' measures of interest (mois).
#'
#' @param dat data.frame to check the existence of rois and mois
#' @param rois,mois vectors of region and measures identifiers, which will be
#' checked against \code{region.col} and \code{measure.col} respectively.
#' The type of entries in the vector should be the same as in the corresponding
#' columns
#' @param region.col,measure.col name of column containing region
#' and eye-tracking measures identifiers respectively. See "Details" for more
#' info.
#'
#' @details
#' The type of \code{region.col} and \code{measure.col} should be either
#' "character" or "quosure" (see documentation for "quosure" in \code{rlang}
#' package)
#' @return Throws an error if some of the rois and mois are not present in the
#' data.frame; the error text indicates which rois/mois are missing.
#'
regions.in.the.data <- dat[[rlang::quo_name(region.col)]]
measures.in.the.data <- dat[[rlang::quo_name(measure.col)]]
if (any(!rois %in% regions.in.the.data)){
stop(paste0("Some rois are not found in the data: ",
paste(rois[!rois %in% regions.in.the.data], collapse = ", ")))
}
if (any(!mois %in% measures.in.the.data)){
stop(paste0("Some measures of interest are not found in the data: ",
paste(mois[!mois %in% measures.in.the.data], collapse = ",")))
}
}
zero_inflate_cells_count <- function(dat, by,
cells.count, rois, mois){
#' Make sure that every roi + moi combination has counts
#'
#' This function ensures that every combination of region and measure
#' contains counts. Basically, it takes care fo the situation when a given
#' region-measure doesn't contain cells of the required kind (e.g. NAs or
#' extreme observations). In such cases this function adds 0 as count to make
#' sure that all region-measure combination are present in the summary plot.
#'
#' @param dat data.frame containing information about \code{by} column
#' @param by quosure indicating which column we are aggregating along. Usually
#' it is the subject or the item column.
#' @param cells.count data.frame with cells count data produced by \code{count_cells}.
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted
#'
#' @return data.frame containing the original cells counts as well as added
#' 0 counts
#' @importFrom dplyr %>%
by <- rlang::quo_name(by)
# if there are no missing cells for some combination of region and measure,
# insert 0s; otherwise that column will be completely missing
# expand.grid varies the first column fastest - this is what we need, but the
# order is wrong: e.g. we want spillover_ff, not, ff_spillover...
reg.meas.combs <- expand.grid(mois, rois)
# so we exchange columns
reg.meas.combs <- reg.meas.combs[c(2,1)]
# then paste them together; mdply adds a new column with the result,
# and we only need that, so we only take the third column
reg.meas.combs <- plyr::mdply(reg.meas.combs, paste, sep="_")[[3]]
# create a dummy data.frame with all necessary region+measure combinations
# filled with 0s
dummy.df <- data.frame(tmp = unique(dat[[by]]),
cell = rep(reg.meas.combs,
each = length(unique(dat[[by]]))),
count = 0,
stringsAsFactors = FALSE)
colnames(dummy.df)[1] <- by
# convert to character, otherwise bind_rows_ will throw a warning.
# It's harmless, but we'd better avoi it not to worry the end user unduly
cells.count$cell <- as.character(cells.count$cell)
cells.count <- dplyr::bind_rows(dummy.df, cells.count) %>%
dplyr::group_by_(by, "cell") %>%
dplyr::summarize(count = sum(count)) %>%
dplyr::ungroup()
cells.count$cell <- factor(cells.count$cell,
levels = reg.meas.combs)
return(cells.count)
}
# ----- Counting functions ----
count_cells <- function(dat, by, ..., cast.formula,
value.name = "count") {
#' Count number of observations of specific kind
#'
#' A generic function to count number of certain observations (e.g. NAs or
#' extreme values) by cells (e.g. combination of region and eye-tracking measure).
#'
#' @param dat data.frame containing data to process
#' @param by unquoted column name for the main grouping variable. See "Details"
#' @param ... unqouted column names - other grouping variables. See "Details".
#' @param cast.formula character. A formula for \code{dcast()} (from 'reshape2' package).
#' LHS is the grouping variable, RHS defines cells
#' @param value.name character. name of the column containing counts in the
#' result data.frame
#'
#' @details
#' The function assumes that \code{dat} argument contains the data which has
#' already been subsetted to only contain values of interest. E.g., if you
#' want to count NAs, before passing the data set to this function, you need
#' to filter all non-NAs out. The package contains two specific convenience
#' functions which would do the subseetting for you (the names are self-explanatory):
#' \code{count_NAs} and \code{count_extremes}.
#'
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which will be displayed on the y axis
#' in the summary plots.
#'
#' The column names in \code{...} argument define which columns will act as
#' grouping variables for \code{dplyr::group_by}, thus defining the smallest
#' subset of the data in which the observations should be counted. In a typical
#' use for eye-tracking data, there will be two components here: region and
#' measure columns names. In general, the values in the first item will be
#' varied slowest, the values in the last item will be varied fastest.
#'
#' Thus, e.g. (if \code{by = subj} and \code{... = region.col, measure,col}),
#' the observations will be counted in each region, in each measure,
#' for each subject.
#'
#'@return
#'
#' data.frame with 3 columns (default names):
#'
#' 1. Name of the \code{by} argument, typically subj or item; contains
#' the unique values from the corresponding column in \code{dat}
#' 2. \code{cell}; contains data subset identifiers. In a typical use,
#' if used with regions and measures, it would be region_measure, e.g. "critical_tt"
#' 3. \code{count} count of observations in each subset of the data.
#'
#'@seealso \code{\link{count_NAs}}, \code{\link{count_extremes}}
#'
#' @importFrom dplyr %>%
#' @importFrom rlang !!
#' @importFrom rlang !!!
grouping_vars <- rlang::list2(...)
id.var <- rlang::quo_name(by)
if (!is.data.frame(dat)){
stop("`dat` argument must be a data.frame")
}
if (!plyr::is.formula(cast.formula)){
stop("`cast.formula` argument must be a formula")
}
if (!is.character(value.name)){
stop("`value.name` should be a character")
}
if (!is.character(id.var)){
stop("`id.var` should be a character")
}
if (nrow(dat) == 0){
warning("Cannot count cells: the data is empty")
res <- data.frame(a = numeric(),
b = character(),
c = numeric())
colnames(res) <- c(id.var, "cell", value.name)
return(res)
}
res <- dat %>%
dplyr::group_by(!!!grouping_vars, !!by) %>%
dplyr::summarize(res =n()) %>%
reshape2::dcast(cast.formula, value.var = "res", fill = 0) %>%
reshape2::melt(id.vars = id.var, variable.name = "cell", value.name = value.name)
return(res)
}
count_NAs <- function(dat, by,
rois, mois,
value.col, region.col, measure.col){
#' Convenience wrapper for count_cells() for counting NAs.
#'
#' Count the number of NAs in the data briken down by region, measure and
#' some third factor, usually subject or item.
#'
#' @param dat data.frame containing data to process
#' @param by quosure. column name for the main grouping variable. See "Details"
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted. Any of these can be left unsepcified, in this
#' case, all regions and/or measure form the data will be used.
#' @param value.col quosure. column name containing the NAs (potentially along
#' with non-NA values. Typically it will be the column containing the reaction
#' times or something alike).
#' @param region.col quosure. column name containing region identifiers
#' @param measure.col quosure. column name containing measure identifiers
#'
#' @details
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which varies slowest and will be
#' displayed on the y axis in the summary plots.
#'
#' @return data.frame with 3 columns analogous to \code{\link{count_cells}}
#' (i.e. containing the main grouping variable identifier (typically subject or item),
#' the combination of region and measure identifying subsets of data, and NAs counts)
#' @export
#' @importFrom dplyr %>%
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
rois <- unique(dat[[rlang::quo_name(measure.col)]])
}
check_columns_existence(dat = dat,
columns = c(rlang::quo_name(by),
rlang::quo_name(value.col),
rlang::quo_name(region.col),
rlang::quo_name(measure.col)))
check_rois_mois_existence (dat = dat, rois = rois, mois = mois,
region.col = region.col,
measure.col = measure.col)
filtered.dat <- dat %>%
dplyr::filter(!stats::complete.cases(rlang::UQ(value.col)))
if (nrow(filtered.dat) == 0){
warning(paste0("There are no NAs in column `", rlang::quo_name(value.col), "`"))
res <- data.frame(a = numeric(),
cell = character(),
count = numeric())
colnames(res)[1] <- rlang::quo_name(by)
} else {
filtered.dat <- filtered.dat%>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois)
res <- count_cells(filtered.dat, by = by,
region.col,
measure.col,
cast.formula = stats::formula(paste(rlang::quo_name(by),"~" ,rlang::quo_name(region.col),
" + ", rlang::quo_name(measure.col))))
}
res <- zero_inflate_cells_count(dat = dat, by = by,
cells.count = res,
rois = rois, mois = mois)
return(res)
}
count_extremes <- function(dat, by,
rois, mois,
value.col, region.col, measure.col,
max.cutoff, min.cutoff = 0){
#' Convenience wrapper for \code{\link{count_cells}} for counting NAs.
#'
#' Count the number of NAs in the data briken down by region, measure and
#' some third factor, usually subject or item.
#'
#' @param dat data.frame containing data to process
#' @param by quosure with column name for the main grouping variable. See "Details"
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted. Any of these can be left unsepcified, in this
#' case, all regions and/or measure form the data will be used.
#' @param value.col quosure with column name containing the NAs (potentially along
#' with non-NA values. Typically it will be the column containing the reaction
#' times or something alike).
#' @param region.col quosure with column name containing region identifiers
#' @param measure.col quosure with column name containing measure identifiers
#' @param max.cutoff numeric value, indicating the upper threshold; any values
#' above this threshold will be considered "extreme" and counted
#' @param min.cutoff numeric value, indicating the lower threshold. Defaults to 0
#'
#' @details
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which varies slowest and will be
#' displayed on the y axis in the summary plots.
#'
#' @return data.frame with 4 columns. The first 3 are analogous to
#' \code{\link{count_cells}}
#' (i.e. they contain the main grouping variable identifier (typically subject or item),
#' the combination of region and measure identifying subsets of data,
#' and extreme values counts), and the fourth one, called "direction", indicates
#' whether the value is extremely high (then the corresponding row will contain
#' "above.max" in the"direction" column) or extremely low ("below.min" in the
#' "direction" column).
#' @export
#' @importFrom dplyr %>%
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
rois <- unique(dat[[rlang::quo_name(measure.col)]])
}
check_columns_existence(dat = dat,
columns = c(rlang::quo_name(by),
rlang::quo_name(value.col),
rlang::quo_name(region.col),
rlang::quo_name(measure.col)))
check_rois_mois_existence (dat = dat, rois = rois, mois = mois,
region.col = region.col,
measure.col = measure.col)
dat <- dat %>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois)
high.extremes <- dat %>%
dplyr::filter(rlang::UQ(value.col) > max.cutoff)
low.extremes <- dat %>%
dplyr::filter(rlang::UQ(value.col) < min.cutoff)
if (nrow(high.extremes) == 0){
warning(paste0("There are no observations higher than ", max.cutoff))
high.extremes <- data.frame(a = numeric(),
cell = character(),
count = numeric(),
direction = character())
colnames(high.extremes)[1] <- rlang::quo_name(by)
} else {
high.extremes <- count_cells(high.extremes,
by,
region.col,
measure.col,
cast.formula = stats::formula(paste(rlang::quo_name(by),"~",
rlang::quo_name(region.col), " + ",
rlang::quo_name(measure.col))))
}
if (nrow(low.extremes) == 0){
warning(paste0("There are no observations lower than ", min.cutoff))
low.extremes <- data.frame(a = numeric(),
cell = character(),
count = numeric(), direction = character())
colnames(low.extremes)[1] <- rlang::quo_name(by)
} else {
low.extremes <- count_cells(low.extremes, by,
region.col,
measure.col,
cast.formula = stats::formula(paste(rlang::quo_name(by),"~",
rlang::quo_name(region.col), " + ",
rlang::quo_name(measure.col))))
}
high.extremes <- zero_inflate_cells_count(dat = dat, by = by,
cells.count = high.extremes,
rois = rois, mois = mois)
low.extremes <- zero_inflate_cells_count(dat = dat, by = by,
cells.count = low.extremes,
rois = rois, mois = mois)
high.extremes$direction <- "above.max"
low.extremes$direction <- "below.min"
res <- rbind(high.extremes, low.extremes)
return(res)
}
# ---- Plotting functions ----
plot_cells_heatmap <- function(dat, by, fill, limits = NULL,
plot.name){
#' Plot heatmap of cell counts
#'
#' A generic function to plot heatmaps of cell counts. Shouldn't be used directly
#' to plot counts produced by \code{\link{count_cells}} and related functions, since
#' those counts may lack the data from certain subsets of the data, if those
#' subsets lack the relevant kind of values (e.g. NAs or extreme values).
#' Instead, \code{\link{plot_cells_count}}, or convenience wrappers around it
#' should be called, since they take care of subsets with 0 observations
#' by artificially adding these 0s back to the counts.
#'
#' @param dat data.frame returned by \code{count_NAs} or \code{count_extremes()}.
#' It should have three columns: grouping variable (subj or item),
#' name of the cell (region+fixationtype by default),
#' number of observations in the cell
#' @param by quosure. name of the grouping variable (normally, first column
#' in \code{dat})
#' @param fill quosure. name of the column w/count of observations (normally,
#' third column in \code{dat})
#' @param limits vector of two integers: lowest and highest N of observations
#' possible in the cell to calibrate color coding. If NULL (default),
#' the limits will be set automatically based on the data.
#' @param plot.name character
#' @return heatmap plot of the cell counts
by <- rlang::quo_name(by)
fill <- rlang::quo_name(fill)
check_columns_existence(dat = dat,
columns = c(by,
fill))
# + scale_y_reverse - make the 1st subject/item appear on top;
# + breaks - make sure that each subject has its number shown
out <- ggplot2::ggplot(dat, ggplot2::aes_string(x = "cell", y = by)) +
ggplot2::geom_tile(ggplot2::aes_string(fill = fill), color = "white") +
ggplot2::scale_fill_gradient(low = "white", high = "steelblue",
breaks = seq(min(limits), max(limits), by = 1)) +
ggplot2::geom_text(ggplot2::aes_string(label = fill))+
ggplot2::scale_x_discrete(position = "top") +
ggplot2::ggtitle (plot.name)
if (is.numeric(dat[[by]])){
out <- out + ggplot2::scale_y_reverse(breaks = sort(unique(dat[[by]])))
}
return(out)
}
plot_cells_count <- function(dat, by,
cells.count, rois, mois, fill = count,
plot.name = "Give a sensible name to the plot!"){
#' Plot heatmaps of cell counts
#'
#' A wrapper around \code{\link{plot_cells_heatmap}} which is aware of regions
#' and measures of interest. Because of that it a) adds 0 counts to region + measure
#' combinations which lacked the relevant kind of value (e.g. extreme or NAs);
#' b) visually delimits ROIs in the plot for visual convenience.
#'
#' @param dat data.frame containing information about \code{by} column
#' @param by quosure indicating which column we are aggregating along. Usually
#' it is the subject or the item column.
#' @param cells.count data.frame with cells count data produced by \code{\link{count_cells}}.
#' The function expects that the data.frame has column "cell", containing
#' the names of data subsets for which counts were performed (e.g
#' region+measure combination) and column "count", containing the counts
#' of the relevant values in each cell.
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted
#' @param fill unquoted name of the column with cell counts
#' @param plot.name character
#'
#' @return heatmap plot of the cell counts (expanded compared to the plot
#' returned by \code{\link{plot_cells_heatmap}} - with added zero counts and ROIs
#' delimitation)
#' @export
if (!rlang::is_quosure(by)) {
stop("`by` argument should be a quosure")
}
check_columns_existence(dat = cells.count,
columns = c("cell",
"count"))
fill <- rlang::enquo(fill)
cells.count$cell <- normalize_axes_labels(cells.count$cell)
# determine the max number for fill scale
fill.upper.limit <- max(cells.count$count)
# How many regions and measures? This is used below in geom_vline() statement
# to figure out, how close should the separating vertical lines inserted
n.mois <- length(mois)
n.rois <- length(rois)
cells.plot <- plot_cells_heatmap(cells.count, by = by, fill = fill,
limits = c(0, fill.upper.limit),
plot.name = plot.name)
if (n.rois > 1){
# Insert separating lines between regions for visual convenience. Start after
# first n + 0.5 measures (+0.5 in order to insert the line in between two columns
# of figures), insert them every n, with an upper limit on n =
# n_of_regions * n_of_measures.
cells.plot <- cells.plot +
ggplot2::geom_vline(xintercept = seq(from = n.mois+0.5, to = n.rois*n.mois, by = n.mois))
}
return(cells.plot)
}
# ---- User-level functions ----
report_NAs_count <- function(dat, by = subj,
rois, mois,
value.col = value,
region.col = reg.name, measure.col = fixationtype,
plot.name = "Give a sensible name to the plot!"){
#' Count and plot NAs by regions and measures of interest
#'
#' @param dat data.frame containing the data where the NAs should be counted
#' @param by unquoted column name for the main grouping variable. See "Details"
#' @param rois,mois vectors of region and measures identifiers, for which counts
#' should be produced. The type of entries in the vector should be the same as in
#' \code{region.col} and \code{measure.col} respectively. Any of these can be left
#' unsepcified, in this case, all regions and/or measure form the data will be used.
#' @param value.col unquoted name of the column containing NAs which should
#' be counted (typically it would be the column containing reaction times
#' or something similar)
#' @param region.col unquoted name of the column containing region ids
#' @param measure.col unquoted name of the column containing measure ids
#' @param plot.name character. name for the plot visualizing the counts
#'
#' @details
#' The \code{by} argument would typically contain subject or item column name.
#' this is the main grouping variable, which varies slowest and will be
#' displayed on the y axis in the summary plots.
#'
#' @return list with the following components:
#' * counts data.frame containing count data
#' * cells.plot heatmap visualization of the counts
#' @export
#' @importFrom dplyr %>%
value.col <- rlang::enquo(value.col)
region.col <- rlang::enquo(region.col)
measure.col <- rlang::enquo(measure.col)
by <- rlang::enquo(by)
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
mois <- unique(dat[[rlang::quo_name(measure.col)]])
}
cells.count <- dat %>%
dplyr::ungroup() %>%
dplyr::do(., count_NAs(.,
rois = rois, mois = mois,
value.col = value.col, region.col = region.col,
measure.col = measure.col,
by = by))
cells.plot <- plot_cells_count(dat = dat,
cells.count = cells.count,
rois = rois, mois = mois,
by = by, plot.name = plot.name)
return(list(counts = cells.count,
plot = cells.plot))
}
report_extremes_count <- function(dat, by = subj,
rois, mois,
value.col = value,
region.col = reg.name, measure.col = fixationtype,
max.cutoff, min.cutoff = 0,
plot.name = "Give a sensible name to the plot!"){
#' Count and plot extreme observation by regions and measures of interest
#'
#' @param dat data.frame containing information about \code{by} column
#' @param by quosure indicating which column we are aggregating along. Usually
#' it is the subject or the item column.
#' @param rois,mois vectors of identifiers for regions and measures, in which
#' cells have to be counted. The type of entries in the vector should be the same as in
#' \code{region.col} and \code{measure.col} respectively.
#' Any of these can be left unsepcified, in this case, all regions and/or measure
#' form the data will be used.
#' @param value.col unquoted name of the column containing NAs which should
#' be counted (typically it would be the column containing reaction times
#' or something similar)
#' @param region.col unquoted name of the column containing region ids
#' @param measure.col unquoted name of the column containing measure ids
#' @param max.cutoff numeric value, indicating the upper threshold; any values
#' above this threshold will be considered "extreme" and counted
#' @param min.cutoff numeric value, indicating the lower threshold. Defaults to 0
#' @param plot.name Name of the heatmap plot with extreme values counts
#'
#' @return list with 3 components:
#' * \code{counts} - data.frame with counts data
#' * \code{count.plots} - list with 3 components
#' * \code{high.plot} - plot with counts of extreme observations above \code{max.cutoff}
#' * \code{low.plot} - plot with counts of extreme observations below \code{min.cutoff}
#' * \code{all.plot} - plot with combined counts of both high and low extreme values
#' * \code{values.plots} - list with 2 components
#' * \code{high.plot} - plot with numerical values of extreme observations above \code{max.cutoff}
#' * \code{low.plot} - plot with numerical values of extreme observations below \code{min.cutoff}
#' @export
#' @importFrom dplyr %>%
value.col <- rlang::enquo(value.col)
region.col <- rlang::enquo(region.col)
measure.col <- rlang::enquo(measure.col)
by <- rlang::enquo(by)
if (missing(rois)){
rois <- unique(dat[[rlang::quo_name(region.col)]])
}
if (missing(mois)){
mois <- unique(dat[[rlang::quo_name(measure.col)]])
}
# count extreme values
cells.count <- dat %>%
dplyr::do(., count_extremes(., by = by,
rois = rois, mois = mois,
value.col = value.col,
region.col = region.col,
measure.col = measure.col,
max.cutoff = max.cutoff,
min.cutoff = min.cutoff))
count.plots <- list()
values.plots <- list()
# plot values above the max cutoff
count.plots$high <- cells.count %>%
dplyr::filter(direction == "above.max") %>%
plot_cells_count(., dat = dat,
rois = rois, mois = mois,
by = by, plot.name = paste0(plot.name, "_above_", max.cutoff))
# plot values below the min cutoff
count.plots$low <- cells.count %>%
dplyr::filter(direction == "below.min") %>%
plot_cells_count(., dat = dat,
rois = rois, mois = mois,
by = by, plot.name = paste0(plot.name, "_below_", min.cutoff))
# combine the two plots above in one
count.plots$all <- cells.count %>%
dplyr::group_by_(by, "cell") %>%
dplyr::summarize(count = sum(count)) %>%
dplyr::ungroup() %>%
plot_cells_count(., dat = dat,
rois = rois, mois = mois,
by = by, plot.name = paste0(plot.name, "_above_", max.cutoff,
"_and_below_", min.cutoff))
# plot numerical values above the max cutoff
values.plots$high <- dat %>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois) %>%
dplyr::filter(rlang::UQ(value.col) > max.cutoff) %>%
ggplot2::ggplot() +
ggplot2::geom_point(ggplot2::aes_string(y = rlang::quo_name(value.col),
x = rlang::quo_name(measure.col)))+
ggplot2::facet_wrap(stats::formula(paste("~", rlang::quo_name(region.col))))
# plot numerical values below the min cutoff
values.plots$low <- dat %>%
dplyr::filter(rlang::UQ(region.col) %in% rois & rlang::UQ(measure.col) %in% mois) %>%
dplyr::filter(rlang::UQ(value.col) < min.cutoff) %>%
ggplot2::ggplot() +
ggplot2::geom_point(ggplot2::aes_string(y = rlang::quo_name(value.col),
x = rlang::quo_name(measure.col)))+
ggplot2::facet_wrap(stats::formula(paste("~", rlang::quo_name(region.col))))
return(list(counts = cells.count,
count.plots = count.plots,
values.plots = values.plots))
}
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.