R/pirateplot_function.R

In ndphillips/yarrr: A Companion to the e-Book "YaRrr!: The Pirate's Guide to R"

#' pirateplot
#'
#' The pirateplot function creates an RDI (Raw data, Descriptive and Inferential statistic) plot showing the relationship between 1 to 3 categorical independent variables and 1 continuous dependent variable.
#'
#' @param formula formula. A formula in the form \code{y ~ x1 + x2 + x3} indicating the vertical response variable (y) and up to three independent variables
#' @param data Either a dataframe containing the variables specified in formula, a list of numeric vectors, or a numeric dataframe / matrix.
#' @param plot logical. If \code{TRUE} (the default), thent the pirateplot is produced. If \code{FALSE}, the data summaries created in the plot are returned as a list.
#' @param pal string. The color palette of the plot. Can be a single color, a vector of colors, or the name of a palette in the piratepal() function (e.g.; "basel", "google", "southpark"). To see all the palettes, run \code{piratepal(palette = "all", action = "show")}
#' @param pointpars dataframe A dataframe containing optional additional plotting parameters for points. The dataframe should have thee same number of rows as \code{data}, and have column names in the set \code{col, bg, pch, labels}.
#' @param mix.col,mix.p Optional color mixing arguments to be passed to \code{piratepal}. See \code{?piratepal} for examples.
#' @param point.col,bar.f.col,bean.b.col,bean.f.col,inf.f.col,inf.b.col,avg.line.col,bar.b.col,quant.col,point.bg string. Vectors of colors specifying the colors of the plotting elements. This will override values in the palette. f stands for filling, b stands for border.
#' @param theme integer. An integer in the set 0, 1, 2 specifying a theme (that is, new default values for opacities and colors). \code{theme = 0} turns off all opacities which can then be individually specified individually.
#' @param beside logical. If FALSE, then the second independent variable is spread over separate plots, rather than separated in the samep lot.
#' @param bar.f.o,point.o,inf.f.o,inf.b.o,avg.line.o,bean.b.o,bean.f.o,bar.b.o numeric. A number between 0 and 1 indicating how opaque to make the bars, points, inference band, average line, and beans respectively. These values override whatever is in the specified theme
#' @param avg.line.fun function. A function that determines how average lines and bar heights are determined (default is mean).
#' @param back.col string. Color of the plotting background.
#' @param point.cex,point.pch,point.lwd numeric.  The size, pch type, and line width of raw data points.
#' @param bean.lwd,bean.lty,inf.lwd,avg.line.lwd,bar.lwd numeric. Vectors of numbers customizing the look of beans and lines.
#' @param width.min,width.max numeric. The minimum/maximum width of the beans.
#' @param cut.min,cut.max numeric. Optional minimum and maximum values of the beans.
#' @param inf.method string. A string indicating what types of inference bands to calculate. "ci" means frequentist confidence intervals, "hdi" means Bayesian Highest Density Intervals (HDI), "iqr" means interquartile range, "sd" means standard deviation, "se" means standard error, "withinci" means frequentist confidence intervals in a within design (Morey, 2008).
#' @param inf.within string. The variable which serves as an ID variable in a within design.
#' @param inf.disp string. How should inference ranges be displayed? \code{"line"} creates a classic vertical line, \code{"rect"} creates a rectangle, \code{"bean"} forms the inference around the bean.
#' @param inf.p numeric. A number adjusting how inference ranges are calculated. for \code{"ci"} and \code{"hdi"}, a number between 0 and 1 indicating the level of confidence (default is .95). For \code{"sd"} and \code{"se"}, the number of standard deviations / standard errors added to or subtracted from the mean (default is 1).
#' @param hdi.iter integer. Number of iterations to run when calculating the HDI. Larger values lead to better estimates, but can be more time consuming.
#' @param bw,adjust Arguments passed to density calculations for beans (see ?density)
#' @param jitter.val numeric. Amount of jitter added to points horizontally. Defaults to 0.05.
#' @param at integer. Locations of the beans. Especially helpful when adding beans to an existing plot with add = TRUE
#' @param sortx string. How to sort the x values. Can be "sequential" (as they are found in the original dataframe), "alphabetical", or a string in the set ("mean", "median", "min", "max") indicating a function
#' @param decreasing logical. If sortx is a named function, should values be sorted in decreasing order?
#' @param add logical. Should the pirateplot elements be added to an existing plotting space?
#' @param cap.beans logical. Should maximum and minimum values of the bean densities be capped at the limits found in the data? Default is FALSE.
#' @param quant.length,quant.lwd numeric. Specifies line lengths/widths of \code{quant}.
#' @param family a font family (Not currently in use)
#' @param cex.lab,cex.axis,cex.names Size of the labels, axes, and bean names.
#' @param force.layout logical. If TRUE, then all internal plotting formatting functions (e.g.; mfrow, mar) are turned off.
#' @param yaxt.y numeric. Locations of tick marks on the y-axis
#' @param gl.y numeric. Locations of the horizontal grid lines
#' @param gl.lwd,gl.lty,gl.col Customization for grid lines. Can be entered as vectors for alternating gridline types
#' @param bty,xlim,ylim,xlab,ylab,main,yaxt,xaxt General plotting arguments
#' @param quant numeric. Adds horizontal lines representing custom quantiles.
#' @param bar.b.lwd,line.fun,inf.o,bean.o,inf.col,theme.o,inf,inf.type,inf.band,bar.o,line.o,hdi.o,gl depricated arguments
#' @keywords plot
#' @importFrom BayesFactor ttestBF
#' @importFrom grDevices col2rgb gray rgb
#' @importFrom graphics abline axis layout mtext par plot points polygon rasterImage rect segments text axTicks
#' @importFrom stats density model.frame optimize rnorm t.test qbeta sd quantile IQR aggregate as.formula
#' @importFrom utils vignette
#' @export
#' @examples
#'
#'
#' # Default pirateplot of weight by Time
#' pirateplot(
#'   formula = weight ~ Time,
#'   data = ChickWeight
#' )
#'
#' # Same but in grayscale
#' pirateplot(
#'   formula = weight ~ Time,
#'   data = ChickWeight,
#'   pal = "gray"
#' )
#'
#'
#' # Now using theme 2
#' pirateplot(
#'   formula = weight ~ Time,
#'   data = ChickWeight,
#'   main = "Chicken weight by time",
#'   theme = 2
#' ) # theme 2
#'
#' # theme 3
#' pirateplot(
#'   formula = weight ~ Time,
#'   data = ChickWeight,
#'   main = "Chicken weight by time",
#'   theme = 3
#' ) # theme 3
#'
#' # theme 4
#' pirateplot(
#'   formula = weight ~ Time,
#'   data = ChickWeight,
#'   main = "Chicken weight by time",
#'   theme = 4
#' ) # theme 4
#'
#' # Start with theme 2, but then customise!
#' pirateplot(
#'   formula = weight ~ Time,
#'   data = ChickWeight,
#'   theme = 2, # theme 2
#'   pal = "xmen", # xmen palette
#'   main = "Chicken weights by Time",
#'   point.o = .4, # Add points
#'   point.col = "black",
#'   point.bg = "white",
#'   point.pch = 21,
#'   bean.f.o = .2, # Turn down bean filling
#'   inf.f.o = .8, # Turn up inf filling
#'   gl.col = "gray", # gridlines
#'   gl.lwd = c(.5, 0)
#' ) # turn off minor grid lines
#'
#' # 2 IVs
#' pirateplot(
#'   formula = len ~ dose + supp,
#'   data = ToothGrowth,
#'   main = "Guinea pig tooth length by supplement",
#'   point.pch = 16, # Point specifications...
#'   point.col = "black",
#'   point.o = .7,
#'   inf.f.o = .9, # inference band opacity
#'   gl.col = "gray"
#' )
#'
#'
#' # Build everything from scratch with theme 0
#' #  And use 3 IVs
#' pirateplot(
#'   formula = height ~ headband + eyepatch + sex,
#'   data = pirates,
#'   pal = gray(.1), # Dark gray palette
#'   theme = 0, # Start from scratch
#'   inf.f.o = .7, # Band opacity
#'   inf.f.col = piratepal("basel"), # Add color to bands
#'   point.o = .1, # Point opacity
#'   avg.line.o = .8, # Average line opacity
#'   gl.col = gray(.6), # Gridline specifications
#'   gl.lty = 1,
#'   gl.lwd = c(.5, 0)
#' )
#'
#' # See the vignette for more details
#' vignette("pirateplot", package = "yarrr")
#'
pirateplot <- function(
    formula = NULL,
    data = NULL,
    plot = TRUE,
    avg.line.fun = mean,
    pal = "basel",
    pointpars = NULL,
    mix.col = "white",
    mix.p = 0,
    back.col = NULL,
    point.cex = NULL,
    point.pch = NULL,
    point.lwd = 1,
    jitter.val = .03,
    theme = 1,
    beside = TRUE,
    bean.b.o = NULL,
    bean.f.o = NULL,
    point.o = NULL,
    bar.f.o = NULL,
    bar.b.o = NULL,
    inf.f.o = NULL,
    inf.b.o = NULL,
    avg.line.o = NULL,
    gl.col = NULL,
    point.col = NULL,
    point.bg = NULL,
    bar.f.col = NULL,
    bean.b.col = NULL,
    bean.f.col = NULL,
    inf.f.col = NULL,
    inf.b.col = NULL,
    avg.line.col = NULL,
    bar.b.col = NULL,
    quant.col = NULL,
    avg.line.lwd = 4,
    bean.lwd = 1,
    bean.lty = 1,
    inf.lwd = NULL,
    bar.lwd = 1,
    at = NULL,
    bw = "nrd0",
    adjust = 1,
    add = FALSE,
    sortx = "alphabetical",
    decreasing = FALSE,
    cex.lab = 1,
    cex.axis = 1,
    cex.names = 1,
    force.layout = FALSE,
    quant = NULL,
    quant.length = NULL,
    quant.lwd = NULL,
    bty = "o",
    cap.beans = TRUE,
    family = NULL,
    inf.method = "hdi",
    inf.within = NULL,
    inf.p = NULL,
    hdi.iter = 1e3,
    inf.disp = NULL,
    cut.min = NULL,
    cut.max = NULL,
    width.min = .3,
    width.max = NA,
    ylim = NULL,
    xlim = NULL,
    xlab = NULL,
    ylab = NULL,
    main = NULL,
    yaxt = NULL,
    xaxt = NULL,
    yaxt.y = NULL,
    gl.y = NULL,
    gl.lwd = NULL,
    gl.lty = NULL,
    bar.b.lwd = NULL,
    line.fun = NULL,
    line.o = NULL,
    inf.o = NULL,
    bean.o = NULL,
    inf.col = NULL,
    theme.o = NULL,
    bar.o = NULL,
    inf = NULL,
    hdi.o = NULL,
    inf.type = NULL,
    inf.band = NULL,
    gl = NULL) {
  #
  #
  #
  # # #
  #   formula = len ~ dose + supp
  #   data = ToothGrowth
  #   plot = TRUE
  #   avg.line.fun = mean
  #   pal = "basel"
  #   back.col = NULL
  #   point.cex = NULL
  #   point.pch = NULL
  #   point.lwd = 1
  #   jitter.val = .03
  #   theme = 1
  #   beside = TRUE
  #   bean.b.o = NULL
  #   bean.f.o = NULL
  #   point.o = NULL
  #   bar.f.o = NULL
  #   bar.b.o = NULL
  #   inf.f.o = NULL
  #   inf.b.o = NULL
  #   force.layout = FALSE
  #   avg.line.o = NULL
  #   gl.col = NULL
  #   point.col = NULL
  #   point.bg = NULL
  #   bar.f.col = NULL
  #   bean.b.col = NULL
  #   bean.f.col = NULL
  #   inf.f.col = NULL
  #   inf.b.col = NULL
  #   avg.line.col = NULL
  #   bar.b.col = NULL
  #   quant.col = NULL
  #   avg.line.lwd = 4
  #   bean.lwd = 1
  #   bean.lty = 1
  #   inf.lwd = NULL
  #   bar.lwd = 1
  #   at = NULL
  #   bw = "nrd0"
  #   adjust = 1
  #   add = FALSE
  #   sortx = "sequential"
  #   decreasing = TRUE
  #   cex.lab = 1
  #   cex.axis = 1
  #   cex.names = 1
  #   quant = NULL
  #   quant.length = NULL
  #   quant.lwd = NULL
  #   bty = "o"
  #   mix.p <- 0
  #   evidence = FALSE
  #   family = NULL
  #   inf.method = "hdi"
  #   inf.p = .95
  #   hdi.iter = 1e3
  #   inf.disp = "line"
  #   cut.min = NULL
  #   cut.max = NULL
  #   width.min = .3
  #   width.max = NA
  #   ylim = NULL
  #   xlim = NULL
  #   xlab = NULL
  #   ylab = NULL
  #   main = NULL
  #   yaxt = NULL
  #   xaxt = NULL
  #   gl.lwd = NULL
  #   gl.lty = NULL
  #   bar.b.lwd = NULL
  #   line.fun = NULL
  #   inf.o = NULL
  #   bean.o = NULL
  #   bar.o = NULL
  #   line.o = NULL
  #   inf.col = NULL
  #   theme.o = NULL
  #   inf = NULL
  #   inf.type = NULL
  #   inf.band = NULL
  #   cap.beans = TRUE
  #   yaxt.y <- NULL
  #   gl.y <- NULL
  #   pointpars = NULL
  # #
  #
  #

  # -----
  #  SETUP
  # ------
  {
    # Check for depricated arguments
    {
      if (is.null(bar.b.lwd) == FALSE) {
        message("bar.b.lwd is depricated. Use bar.lwd instead")

        bar.lwd <- bar.b.lwd
      }

      if (is.null(line.fun) == FALSE) {
        message("line.fun is depricated. Use avg.line.fun instead")

        avg.line.fun <- line.fun
      }

      if (is.null(inf.o) == FALSE) {
        message("inf.o is depricated. Use inf.f.o instead")

        inf.f.o <- inf.o
      }

      if (is.null(line.o) == FALSE) {
        message("line.o is depricated. Use avg.line.o instead")

        avg.line.o <- line.o
      }


      if (is.null(bean.o) == FALSE) {
        message("bean.o is depricated. Use bean.b.o instead")

        bean.b.o <- bean.o
      }

      if (is.null(inf.col) == FALSE) {
        message("inf.col is depricated. Use inf.f.col instead")

        inf.f.col <- inf.col
      }

      if (is.null(theme.o) == FALSE) {
        message("theme.o is depricated. Use theme instead")

        theme <- theme.o
      }

      if (is.null(inf) == FALSE) {
        message("inf is depricated. Use inf.method instead")

        inf.method <- inf
      }

      if (is.null(inf.band) == FALSE) {
        message("inf.band is depricated. Use inf.disp instead")

        inf.disp <- inf.band
      }

      if (is.null(bar.o) == FALSE) {
        message("bar.o is depricated. Use bar.f.o (for filling), and bar.b.o (for border) instead")

        bar.f.o <- bar.o
      }
    }

    # Look for missing critical inputs
    {
      if (is.null(data)) {
        stop("You must specify data in the data argument")
      }
    }

    # Set some defaults

    if (inf.method %in% c("hdi", "ci", "withinci") & is.null(inf.p)) {
      inf.p <- .95
    }

    if (inf.method %in% c("sd", "se") & is.null(inf.p)) {
      inf.p <- 1
    }


    # If no formula, than reshape data
    if (is.null(formula)) {
      if (inherits(data, c("data.frame", "matrix"))) {
        # data <- data.frame(a = rnorm(100), b = rnorm(100), c = rnorm(100))
        # data <- matrix(rnorm(100), nrow = 20, ncol = 5)
        #

        data <- as.data.frame(data)
        iv.levels <- names(data)

        data <- stats::reshape(data, direction = "long", varying = list(1:ncol(data)))
        data <- data[, 1:2]
        names(data) <- c("group", "y")

        group.num <- data$group

        for (i in 1:length(iv.levels)) {
          data$group[group.num == i] <- iv.levels[i]
        }


        formula <- y ~ group
      }

      if (inherits(data, c("numeric", "integer"))) {
        data <- data.frame(y = data, group = rep(1, length(data)))

        formula <- y ~ group
        xlab <- ""
        if (is.null(xaxt)) {
          xaxt <- "n"
        }
        if (is.null(ylab)) {
          ylab <- ""
        }
      }






      if (inherits(data, "list")) {
        if (is.null(names(data))) {
          names(data) <- paste0("V", 1:length(data))
        }

        iv.levels <- names(data)

        # Convert list to dataframe

        data.df <- do.call("rbind", lapply(1:length(data), FUN = function(x) {
          data.frame(
            group = rep(iv.levels[x], length(data[[x]])),
            y = unlist(data[[x]])
          )
        }))

        data <- data.df
        formula <- y ~ group
      }
    }

    # Reshape dataframe to include relevant variables
    {
      withindata <- data
      data <- model.frame(
        formula = formula,
        data = data
      )

      dv.name <- names(data)[1]
      dv.v <- data[, 1]
      all.iv.names <- names(data)[2:ncol(data)]
    }



    # GET IV INFORMATION
    {
      n.iv <- ncol(data) - 1
      if (n.iv > 3) {
        stop("Currently only 1, 2, or 3 IVs are supported in pirateplot(). Please reduce.")
      }

      # selection.mtx dictates which values are in each sub-plot
      if (n.iv == 1 | (n.iv == 2 & beside == TRUE)) {
        selection.mtx <- matrix(TRUE, nrow = nrow(data), ncol = 1)
      }

      if (n.iv == 2 & beside == FALSE) {
        iv2.name <- names(data)[3]
        iv2.levels <- sort(unique(data[, 3]))
        selection.mtx <- matrix(unlist(lapply(iv2.levels, FUN = function(x) {
          data[, 3] == x
        })), nrow = nrow(data), ncol = length(iv2.levels), byrow = FALSE)
      }

      if (n.iv == 3) {
        iv3.name <- names(data)[4]
        iv3.levels <- sort(unique(data[, 4]))
        selection.mtx <- matrix(unlist(lapply(iv3.levels, FUN = function(x) {
          data[, 4] == x
        })), nrow = nrow(data), ncol = length(iv3.levels), byrow = F)
      }

      n.subplots <- ncol(selection.mtx)

      # Loop over subplots (only relevant when there is a third IV)
      if (force.layout == FALSE) {
        if (n.subplots == 2) {
          par(mfrow = c(1, 2))
        }
        if (n.subplots == 3) {
          par(mfrow = c(1, 3))
        }
        if (n.subplots == 4) {
          par(mfrow = c(2, 2))
        }
        if (n.subplots %in% c(5, 6)) {
          par(mfrow = c(2, 3))
        }
        if (n.subplots > 7) {
          par(mfrow = c(ceiling(sqrt(n.subplots)), ceiling(sqrt(n.subplots))))
        }
      }
    }

    # Setup outputs
    summary.ls <- vector("list", length = n.subplots)
  }

  # Loop over subplots
  for (subplot.i in 1:n.subplots) {
    # Select data for current subplot
    data.i <- data[selection.mtx[, subplot.i], ]


    if (is.null(pointpars) == FALSE) {
      pointpars.i <- pointpars[selection.mtx[, subplot.i], ]
    }

    # Remove potential iv3 column
    data.i <- data.i[, 1:min(ncol(data.i), 3)]

    # Determine levels of each IV
    {
      if ((substr(sortx, 1, 1) %in% c("a", "s") |
        sortx %in% c("mean", "median", "min", "max")) == FALSE
      ) {
        stop("sortx argument is invalid. use 'alphabetical', 'sequential', 'mean', 'median', 'min', or 'max'")
      }

      if (substr(sortx, 1, 1) == "a") {
        iv.levels <- lapply(2:ncol(data.i),
          FUN = function(x) {
            sort(unique(data.i[, x]))
          }
        )
      }

      if (substr(sortx, 1, 1) == "s") {
        iv.levels <- lapply(2:ncol(data.i), FUN = function(x) {
          unique(data.i[, x])
        })
      }

      if (sortx %in% c("mean", "median", "min", "max")) {
        agg <- aggregate(formula, data = data.i, FUN = get(sortx))


        if (ncol(agg) == 2) {
          agg <- agg[order(agg[, 2], decreasing = decreasing), ]
          iv.levels <- list(paste(agg[, 1]))
        }

        if (ncol(agg) == 3) {
          agg.1 <- aggregate(as.formula(paste(dv.name, "~", all.iv.names[1])),
            data = data.i, FUN = get(sortx)
          )

          agg.1 <- agg.1[order(agg.1[, 2], decreasing = decreasing), ]
          iv.levels <- list(paste(agg.1[, 1]))

          agg.2 <- aggregate(as.formula(paste(dv.name, "~", all.iv.names[2])),
            data = data.i, FUN = get(sortx)
          )

          agg.2 <- agg.2[order(agg.2[, 2], decreasing = decreasing), ]
          iv.levels[2] <- list(paste(agg.2[, 1]))
        }
      }


      iv.lengths <- sapply(1:length(iv.levels), FUN = function(x) {
        length(iv.levels[[x]])
      })
      iv.names <- names(data.i)[2:ncol(data.i)]
      subplot.n.iv <- length(iv.levels)
    }

    # Set up bean info
    {
      bean.mtx <- expand.grid(iv.levels)
      names(bean.mtx) <- names(data.i)[2:ncol(data.i)]
      n.beans <- nrow(bean.mtx)
      bean.mtx$bean.num <- 1:nrow(bean.mtx)

      # Determine bean x locations

      if (is.null(at)) {
        bean.loc <- 1:n.beans

        group.spacing <- 1

        if (subplot.n.iv == 2) {
          bean.loc <- bean.loc + rep(group.spacing * (0:(iv.lengths[2] - 1)),
            each = iv.lengths[1]
          )
        }
      }

      if (!is.null(at)) {
        bean.loc <- rep(at, length.out = n.beans)
      }

      bean.mtx$x.loc <- bean.loc
      data.i <- merge(data.i, bean.mtx)
    }

    # Calculate summary statistics for each bean
    {
      summary <- data.frame(
        "n" = rep(NA, n.beans),
        "avg" = rep(NA, n.beans),
        "inf.lb" = rep(NA, n.beans),
        "inf.ub" = rep(NA, n.beans)
      )

      summary <- cbind(bean.mtx[, (1:ncol(bean.mtx) - 1)], summary)

      if (n.subplots > 1 & beside == TRUE) {
        summary[iv3.name] <- iv3.levels[subplot.i]
      }


      # Loop over beans in subplot
      for (bean.i in 1:n.beans) {
        dv.i <- data.i[data.i$bean.num == bean.i, dv.name]
        if (is.logical(dv.i)) {
          dv.i <- as.numeric(dv.i)
        }

        summary$n[bean.i] <- length(dv.i)
        summary$avg[bean.i] <- avg.line.fun(dv.i)

        # Calculate inference

        if (length(dv.i) > 0) {
          # Binary data.i

          if (length(setdiff(dv.i, c(0, 1))) == 0) {
            if (inf.method == "hdi") {
              # Calculate HDI from beta(Success + 1, Failure + 1)
              inf.lb <- qbeta(.025, shape1 = sum(dv.i) + 1, shape2 = sum(dv.i == 0) + 1)
              inf.ub <- qbeta(.975, shape1 = sum(dv.i) + 1, shape2 = sum(dv.i == 0) + 1)
            }

            if (inf.method == "ci") {
              # Calculate 95% CI with Normal distribution approximation to binomial
              inf.lb <- mean(dv.i) - 1.96 * sqrt(mean(dv.i) * (1 - mean(dv.i)) / length(dv.i)) - .5 / length(dv.i)
              inf.ub <- mean(dv.i) + 1.96 * sqrt(mean(dv.i) * (1 - mean(dv.i)) / length(dv.i)) + .5 / length(dv.i)

              if (inf.lb < 0) {
                inf.lb <- 0
              }
              if (inf.lb > 1) {
                inf.ub <- 1
              }
            }

            if (inf.method == "sd") {
              inf.lb <- mean(dv.i) - sd(dv.i) * inf.p
              inf.ub <- mean(dv.i) + sd(dv.i) * inf.p

              if (inf.lb < 0) {
                inf.lb <- 0
              }
              if (inf.lb > 1) {
                inf.ub <- 1
              }
            }

            if (inf.method == "se") {
              inf.lb <- mean(dv.i) - sd(dv.i) / sqrt(length(dv.i)) * inf.p
              inf.ub <- mean(dv.i) + sd(dv.i) / sqrt(length(dv.i)) * inf.p

              if (inf.lb < 0) {
                inf.lb <- 0
              }
              if (inf.lb > 1) {
                inf.ub <- 1
              }
            }
          }

          # Non-Binary data.i
          if (length(setdiff(dv.i, c(0, 1))) > 0) {
            if (inf.method == "hdi") {
              ttest.bf <- BayesFactor::ttestBF(dv.i, posterior = T, iterations = hdi.iter, progress = F)
              samples <- ttest.bf[, 1]

              # using the hdi function from Kruschke

              inf.lb <- hdi(samples, credMass = inf.p)[1]
              inf.ub <- hdi(samples, credMass = inf.p)[2]
            }

            if (inf.method == "iqr") {
              inf.lb <- quantile(dv.i, probs = .25)
              inf.ub <- quantile(dv.i, probs = .75)
            }

            if (inf.method == "ci") {
              ci.i <- t.test(dv.i, conf.level = inf.p)$conf.int

              inf.lb <- ci.i[1]
              inf.ub <- ci.i[2]
            }

            if (inf.method == "sd") {
              inf.lb <- mean(dv.i) - sd(dv.i) * inf.p
              inf.ub <- mean(dv.i) + sd(dv.i) * inf.p
            }

            if (inf.method == "se") {
              inf.lb <- mean(dv.i) - sd(dv.i) / sqrt(length(dv.i)) * inf.p
              inf.ub <- mean(dv.i) + sd(dv.i) / sqrt(length(dv.i)) * inf.p
            }

            if (inf.method == "withinci") {
              grandMean <- mean(dv.v)
              Groups <- unique(withindata[inf.within])[[1]]
              groupMean <- c()
              dv.within <- c()

              for (group in 1:length(Groups)) {
                # get the participant/group mean for each participant/group
                groupMean[group] <- mean(dv.v[which(withindata[inf.within] == Groups[group])])
              }

              for (datum in 1:length(dv.i)) {
                # substitute group mean with grand mean, which removes between subject variance
                dv.within[datum] <- dv.i[datum] - groupMean[datum] + grandMean
              }

              ci.i <- t.test(dv.within, conf.level = inf.p)$conf.int

              ci.width <- (ci.i[2] - ci.i[1])
              inf.lb <- mean(dv.i) - (ci.width / 2) * sqrt(n.beans / (n.beans - 1)) # with Morey correction
              inf.ub <- mean(dv.i) + (ci.width / 2) * sqrt(n.beans / (n.beans - 1)) # with Morey correction
            }
          }

          summary$inf.lb[bean.i] <- inf.lb
          summary$inf.ub[bean.i] <- inf.ub
        }

        if (length(dv.i) == 0) {
          summary$inf.lb[bean.i] <- NA
          summary$inf.ub[bean.i] <- NA
        }
      }
    }

    if (plot == TRUE) {
      # COLORS AND TRANSPARENCIES
      {
        # Set number of colors to number of levels of the first IV
        n.cols <- iv.lengths[1]

        # DEFINE THEMES
        {
          if ((theme %in% 0:4) == FALSE) {
            print("theme must be an integer between 0 and 4. I'll set it to 1 for now.")
            theme <- 1
          }

          if (theme == 0) {
            if (is.null(point.pch)) {
              point.pch <- 16
            }
            if (is.null(point.o)) {
              point.o <- 0
            }
            if (is.null(bean.b.o)) {
              bean.b.o <- 0
            }
            if (is.null(bean.f.o)) {
              bean.f.o <- 0
            }
            if (is.null(inf.f.o)) {
              inf.f.o <- 0
            }
            if (is.null(inf.b.o)) {
              inf.b.o <- 0
            }
            if (is.null(avg.line.o)) {
              avg.line.o <- 0
            }
            if (is.null(bar.f.o)) {
              bar.f.o <- 0
            }
            if (is.null(bar.b.o)) {
              bar.b.o <- 0
            }
            if (is.null(point.cex)) {
              point.cex <- 1
            }
            if (is.null(gl.col)) {
              gl.col <- "white"
            }

            if (is.null(inf.disp)) {
              inf.disp <- "rect"
            }
          }

          if (theme == 1) {
            if (is.null(point.o)) {
              point.o <- .2
            }
            if (is.null(bean.b.o)) {
              bean.b.o <- 1
            }
            if (is.null(bean.f.o)) {
              bean.f.o <- .3
            }
            if (is.null(inf.f.o)) {
              inf.f.o <- .8
            }
            if (is.null(inf.b.o)) {
              inf.b.o <- .8
            }
            if (is.null(avg.line.o)) {
              avg.line.o <- 1
            }
            if (is.null(bar.f.o)) {
              bar.f.o <- 0
            }
            if (is.null(bar.b.o)) {
              bar.b.o <- 0
            }
            if (is.null(bean.b.col)) {
              bean.b.col <- "black"
            }
            if (is.null(point.cex)) {
              point.cex <- .7
            }
            if (is.null(point.col)) {
              point.col <- "black"
            }
            if (is.null(inf.b.col)) {
              inf.b.col <- gray(0)
            }
            if (is.null(inf.f.col)) {
              inf.f.col <- "white"
            }
            if (is.null(bean.lwd)) {
              bean.lwd <- 2
            }
            if (is.null(avg.line.col)) {
              avg.line.col <- "black"
            }


            if (is.null(gl.col)) {
              gl.col <- "gray"
            }
            if (is.null(gl.lwd)) {
              gl.lwd <- c(.25, .5)
            }

            if (is.null(point.col)) {
              point.col <- "black"
            }
            if (is.null(point.bg)) {
              point.bg <- "white"
            }
            if (is.null(point.pch)) {
              point.pch <- 16
            }

            if (is.null(inf.disp)) {
              inf.disp <- "rect"
            }
          }

          if (theme == 2) {
            if (is.null(point.pch)) {
              point.pch <- 16
            }
            if (is.null(point.o)) {
              point.o <- .1
            }
            if (is.null(bean.b.o)) {
              bean.b.o <- 1
            }
            if (is.null(bean.f.o)) {
              bean.f.o <- 1
            }
            if (is.null(inf.f.o)) {
              inf.f.o <- .6
            }
            if (is.null(inf.b.o)) {
              inf.b.o <- .8
            }
            if (is.null(inf.b.col)) {
              inf.b.col <- gray(0)
            }
            if (is.null(avg.line.o)) {
              avg.line.o <- 1
            }
            if (is.null(bar.f.o)) {
              bar.f.o <- 0
            }
            if (is.null(bar.b.o)) {
              bar.b.o <- 0
            }
            if (is.null(bean.b.col)) {
              bean.b.col <- "black"
            }
            if (is.null(point.cex)) {
              point.cex <- .7
            }
            if (is.null(point.col)) {
              point.col <- "black"
            }
            if (is.null(bean.lwd)) {
              bean.lwd <- 2
            }
            if (is.null(avg.line.col)) {
              avg.line.col <- "black"
            }
            if (is.null(bean.f.col)) {
              bean.f.col <- "white"
            }

            if (is.null(inf.disp)) {
              inf.disp <- "rect"
            }

            if (is.null(gl.col)) {
              gl.col <- "gray"
            }
            if (is.null(gl.lwd)) {
              gl.lwd <- c(.25, .5)
            }
          }

          if (theme == 3) {
            if (is.null(point.pch)) {
              point.pch <- 16
            }
            if (is.null(point.o)) {
              point.o <- .3
            }
            if (is.null(bean.b.o)) {
              bean.b.o <- 1
            }
            if (is.null(bean.f.o)) {
              bean.f.o <- .5
            }
            if (is.null(inf.f.o)) {
              inf.f.o <- .9
            }
            if (is.null(inf.b.o)) {
              inf.b.o <- 1
            }
            if (is.null(avg.line.o)) {
              avg.line.o <- 1
            }
            if (is.null(bar.f.o)) {
              bar.f.o <- 0
            }
            if (is.null(bar.b.o)) {
              bar.b.o <- 0
            }

            if (is.null(bean.b.col)) {
              bean.b.col <- "black"
            }
            if (is.null(inf.f.col)) {
              inf.f.col <- "white"
            }
            if (is.null(inf.b.col)) {
              inf.b.col <- gray(0)
            }
            if (is.null(avg.line.col)) {
              avg.line.col <- "black"
            }
            if (is.null(point.col)) {
              point.col <- "black"
            }
            if (is.null(point.cex)) {
              point.cex <- .5
            }
            #
            #  if(is.null(back.col)) {back.col <- gray(.97)}
            if (is.null(inf.disp)) {
              inf.disp <- "bean"
            }

            if (is.null(gl.col)) {
              gl.col <- "gray"
            }
            if (is.null(gl.lwd)) {
              gl.lwd <- c(.25, .5)
            }
          }

          if (theme == 4) {
            if (is.null(point.pch)) {
              point.pch <- 16
            }
            if (is.null(point.o)) {
              point.o <- .3
            }
            if (is.null(bean.b.o)) {
              bean.b.o <- 0
            }
            if (is.null(bean.f.o)) {
              bean.f.o <- 0
            }
            if (is.null(inf.f.o)) {
              inf.f.o <- 1
            }
            if (is.null(inf.b.o)) {
              inf.b.o <- .5
            }
            if (is.null(avg.line.o)) {
              avg.line.o <- 1
            }
            if (is.null(bar.f.o)) {
              bar.f.o <- 1
            }
            if (is.null(bar.b.o)) {
              bar.b.o <- 1
            }

            if (is.null(inf.f.col)) {
              inf.f.col <- "black"
            }
            if (is.null(inf.b.col)) {
              inf.b.col <- gray(0)
            }
            if (is.null(avg.line.col)) {
              avg.line.col <- "black"
            }
            if (is.null(bar.f.col)) {
              bar.f.col <- "white"
            }
            if (is.null(bar.b.col)) {
              bar.b.col <- "black"
            }

            if (is.null(point.col)) {
              point.col <- "black"
            }
            if (is.null(point.cex)) {
              point.cex <- .5
            }
            #
            #  if(is.null(back.col)) {back.col <- gray(.97)}

            if (is.null(gl.col)) {
              gl.col <- "gray"
            }
            if (is.null(gl.lwd)) {
              gl.lwd <- c(.25, .5)
            }

            if (is.null(inf.disp)) {
              inf.disp <- "line"
            }
          }


          # Inference lwd depends on inf.disp...
          if (is.null(inf.lwd)) {
            if (inf.disp == "line") {
              inf.lwd <- 3
            }
            if (inf.disp %in% c("rect", "bean")) {
              inf.lwd <- 2
            }
          }
        }

        # DEFINE OPACITIES
        {
          opac.df <- data.frame(
            point.o = rep(NA, n.beans),
            bean.b.o = rep(NA, n.beans),
            bean.f.o = rep(NA, n.beans),
            inf.f.o = rep(NA, n.beans),
            inf.b.o = rep(NA, n.beans),
            avg.line.o = rep(NA, n.beans),
            bar.f.o = rep(NA, n.beans),
            bar.b.o = rep(NA, n.beans)
          )
          rownames(opac.df) <- 1:n.beans




          # If opacity values are specified, update them.
          if (is.null(point.o) == FALSE) {
            opac.df$point.o <- rep(point.o, length.out = n.beans)
          }
          if (is.null(bean.b.o) == FALSE) {
            opac.df$bean.b.o <- rep(bean.b.o, length.out = n.beans)
          }
          if (is.null(bean.f.o) == FALSE) {
            opac.df$bean.f.o <- rep(bean.f.o, length.out = n.beans)
          }
          if (is.null(inf.f.o) == FALSE) {
            opac.df$inf.f.o <- rep(inf.f.o, length.out = n.beans)
          }
          if (is.null(inf.b.o) == FALSE) {
            opac.df$inf.b.o <- rep(inf.b.o, length.out = n.beans)
          }
          if (is.null(avg.line.o) == FALSE) {
            opac.df$avg.line.o <- rep(avg.line.o, length.out = n.beans)
          }
          if (is.null(bar.f.o) == FALSE) {
            opac.df$bar.f.o <- rep(bar.f.o, length.out = n.beans)
          }
          if (is.null(bar.b.o) == FALSE) {
            opac.df$bar.b.o <- rep(bar.b.o, length.out = n.beans)
          }
        }

        # DEFINE COLORS
        {
          colors.df <- data.frame(
            point.col = rep(NA, n.beans),
            point.bg = rep(NA, n.beans),
            bean.b.col = rep(NA, n.beans),
            bean.f.col = rep(NA, n.beans),
            inf.f.col = rep(NA, n.beans),
            inf.b.col = rep(NA, n.beans),
            avg.line.col = rep(NA, n.beans),
            bar.f.col = rep(NA, n.beans),
            bar.b.col = rep(NA, n.beans),
            quant.col = rep(NA, n.beans)
          )
          rownames(colors.df) <- 1:n.beans


          # If palette is in piratepal()...
          if (mean(pal %in% piratepal("names")) == 1) {
            colors.df$point.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$point.bg <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$bean.b.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$bean.f.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$inf.f.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$inf.b.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$avg.line.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$bar.f.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$bar.b.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
            colors.df$quant.col <- rep(piratepal(palette = pal, length.out = n.cols, mix.col = mix.col, mix.p = mix.p), length.out = n.beans)
          }

          # If palette is NOT in piratepal()...

          if (mean(pal %in% piratepal("names")) != 1) {
            colors.df$point.col <- rep(pal, length.out = n.beans)
            colors.df$point.bg <- rep(pal, length.out = n.beans)
            colors.df$bean.b.col <- rep(pal, length.out = n.beans)
            colors.df$bean.f.col <- rep(pal, length.out = n.beans)
            colors.df$inf.f.col <- rep(pal, length.out = n.beans)
            colors.df$inf.b.col <- rep(pal, length.out = n.beans)
            colors.df$avg.line.col <- rep(pal, length.out = n.beans)
            colors.df$bar.f.col <- rep(pal, length.out = n.beans)
            colors.df$bar.b.col <- rep(pal, length.out = n.beans)
            colors.df$quant.col <- rep(pal, length.out = n.beans)
          }

          # Apply specified colors

          if (is.null(point.col) == FALSE) {
            colors.df$point.col <- rep(point.col, length.out = n.beans)
          }
          if (is.null(point.bg) == FALSE) {
            colors.df$point.bg <- rep(point.bg, length.out = n.beans)
          }
          if (is.null(bean.b.col) == FALSE) {
            colors.df$bean.b.col <- rep(bean.b.col, length.out = n.beans)
          }
          if (is.null(bean.f.col) == FALSE) {
            colors.df$bean.f.col <- rep(bean.f.col, length.out = n.beans)
          }
          if (is.null(inf.f.col) == FALSE) {
            colors.df$inf.f.col <- rep(inf.f.col, length.out = n.beans)
          }
          if (is.null(inf.b.col) == FALSE) {
            colors.df$inf.b.col <- rep(inf.b.col, length.out = n.beans)
          }
          if (is.null(avg.line.col) == FALSE) {
            colors.df$avg.line.col <- rep(avg.line.col, length.out = n.beans)
          }
          if (is.null(bar.f.col) == FALSE) {
            colors.df$bar.f.col <- rep(bar.f.col, length.out = n.beans)
          }
          if (is.null(bar.b.col) == FALSE) {
            colors.df$bar.b.col <- rep(bar.b.col, length.out = n.beans)
          }
          if (is.null(quant.col) == FALSE) {
            colors.df$quant.col <- rep(quant.col, length.out = n.beans)
          }
        }
      }

      # SETUP PLOTTING SPACE
      {
        if (force.layout == FALSE) {
          # Determine margins
          if (n.subplots == 1) {
            # par(mar = c(5, 4, 4, 1) + .1)
            add.yaxt <- TRUE
          }
          if (n.subplots %in% c(2, 3)) {
            if (subplot.i == 1) {
              par(mar = c(5, 4, 4, 1) + .1)
              add.yaxt <- TRUE
            }

            if (subplot.i > 1) {
              par(mar = c(5, 3, 4, 1) + .1)
              add.yaxt <- FALSE
            }
          }
          if (n.subplots == c(4)) {
            if (subplot.i %in% c(1, 3)) {
              par(mar = c(5, 4, 4, 1) + .1)
            } else {
              par(mar = c(5, 3, 4, 1) + .1)
            }
          }
          if (n.subplots %in% c(5, 6)) {
            if (subplot.i %in% c(1, 4)) {
              par(mar = c(5, 4, 4, 1) + .1)
            } else {
              par(mar = c(5, 3, 4, 1) + .1)
            }
          }
          if (n.subplots > 6) {
            if (subplot.i %in% seq(ceiling(sqrt(n.subplots)) + 1,
              n.subplots + 1,
              length.out = ceiling(sqrt(n.subplots))
            )) {
              par(mar = c(5, 4, 4, 1) + .1)
            } else {
              par(mar = c(5, 3, 4, 1) + .1)
            }
          }
        }

        # Determine y limits (y axis limits)
        # y axis breaks (yaxt.y)


        if (is.null(ylim) == TRUE) {
          # Determine best step size

          steps.p <- c(
            seq(1e-3, 1e-2, 1e-3),
            seq(1e-4, 1e-3, 1e-3),
            seq(1e-5, 1e-4, 1e-4),
            seq(1e-6, 1e-5, 1e-5),
            seq(1e-7, 1e-6, 1e-6),
            seq(1e-8, 1e-7, 1e-7),
            seq(1e-9, 1e-8, 1e-8),
            1 / 2, 1 / 5, 1 / 10, 1 / 25, 1 / 50, 1 / 100,
            1, 2, 5, 10, 25, 50, 100,
            seq(1e2, 1e3, 1e2),
            seq(1e3, 1e4, 1e3),
            seq(1e4, 1e5, 1e4),
            seq(1e5, 1e6, 1e5),
            seq(1e6, 1e7, 1e6),
            seq(1e7, 1e8, 1e7),
            seq(1e8, 1e9, 1e8)
          )


          range <- max(dv.v) - min(dv.v)

          steps.p.m <- range / steps.p
          best.step.size <- min(steps.p[which(abs(steps.p.m - 10) == min(abs(steps.p.m - 10)))])

          plot.min <- floor(min(dv.v) / best.step.size) * best.step.size
          plot.max <- ceiling(max(dv.v) / best.step.size) * best.step.size
          plot.height <- plot.max - plot.min

          ylim <- c(plot.min, plot.max)



          # yaxt.y <- seq(plot.min, plot.max, by = best.step.size)
        }
        #
        # if(is.null(ylim) == FALSE) {
        #
        #   steps.p <- c(
        #     seq(1e-3, 1e-2, 1e-3),
        #     seq(1e-4, 1e-3, 1e-3),
        #     seq(1e-5, 1e-4, 1e-4),
        #     seq(1e-6, 1e-5, 1e-5),
        #     seq(1e-7, 1e-6, 1e-6),
        #     seq(1e-8, 1e-7, 1e-7),
        #     seq(1e-9, 1e-8, 1e-8),
        #     1/2, 1/5, 1/10, 1/25, 1/50, 1/100,
        #     1, 2, 5, 10, 25, 50, 100,
        #     seq(1e2, 1e3, 1e2),
        #     seq(1e3, 1e4, 1e3),
        #     seq(1e4, 1e5, 1e4),
        #     seq(1e5, 1e6, 1e5),
        #     seq(1e6, 1e7, 1e6),
        #     seq(1e7, 1e8, 1e7),
        #     seq(1e8, 1e9, 1e8)
        #   )
        #
        #   range <- ylim[2] - ylim[1]
        #
        #   steps.p.m <- range / steps.p
        #   best.step.size <- min(steps.p[which(abs(steps.p.m - 10) == min(abs(steps.p.m - 10)))])
        #
        #   plot.min <- floor(ylim[1] / best.step.size) * best.step.size
        #   plot.max <- ceiling((max(dv.v) - plot.min)/  best.step.size) * best.step.size
        #
        #   yaxt.y <- seq(ylim[1], ylim[2], by = best.step.size)
        #
        # }

        # Determine x and y labels

        if (subplot.n.iv == 1 & is.null(xlab)) {
          my.xlab <- iv.names[1]
        }
        if (subplot.n.iv == 1 & is.null(xlab) == FALSE) {
          my.xlab <- xlab
        }

        if (subplot.n.iv > 1) {
          my.xlab <- ""
        }

        if (is.null(ylab)) {
          ylab <- dv.name
        }
      }


      # PLOTTING SPACE
      if (add == FALSE) {
        if (is.null(xlim)) {
          xlim <- c(min(bean.loc) - .5, max(bean.loc) + .5)
        }

        plot(1,
          xlim = xlim,
          ylim = ylim,
          type = "n",
          xaxt = "n",
          yaxt = "n",
          xlab = my.xlab,
          ylab = ylab,
          cex.lab = cex.lab,
          main = main,
          yaxt = yaxt,
          bty = bty # ,
          #   ...
        )

        # Get axis locations

        if (is.null(yaxt.y)) {
          yaxt.y <- axTicks(2)
        }

        # Add title for iv3

        if (n.iv > 2) {
          top.text <- paste(names(data)[4], " = ", iv3.levels[subplot.i], sep = "")
          mtext(text = top.text, side = 3, line = 1)
        }



        # Y-AXIS
        {
          if (is.null(yaxt)) {
            axis(
              side = 2,
              at = yaxt.y,
              labels = prettyNum(yaxt.y, big.mark = ","),
              las = 1,
              lwd = 0,
              lwd.ticks = 1,
              cex.axis = cex.axis
            )
          }
        }

        # BACKGROUND
        {
          if (is.null(back.col) == FALSE) {
            rect(
              xleft = par("usr")[1],
              ybottom = par("usr")[3],
              xright = par("usr")[2],
              ytop = par("usr")[4],
              col = back.col,
              border = NA
            )
          }
        }

        # GRIDLINES
        {
          if (is.null(gl.col) == FALSE) {
            if (is.null(gl.lwd)) {
              gl.lwd <- c(.5)
            }
            if (is.null(gl.lty)) {
              gl.lty <- 1
            }
            if (is.null(gl.y)) {
              gl.y <- yaxt.y
            }

            abline(
              h = gl.y,
              lwd = gl.lwd,
              col = gl.col,
              lty = gl.lty
            )
          }
        }
      }

      # PIRATEPLOT ELEMENTS
      {
        if (is.na(width.max)) {
          if (subplot.n.iv == 1) {
            width.max <- .45
          }
          if (subplot.n.iv == 2) {
            width.max <- .5
          }
        }

        bean.lwd <- rep(bean.lwd, length.out = n.beans)
        bean.lty <- rep(bean.lty, length.out = n.beans)
        inf.lwd <- rep(inf.lwd, length.out = n.beans)
        avg.line.lwd <- rep(avg.line.lwd, length.out = n.beans)
        bar.lwd <- rep(bar.lwd, length.out = n.beans)

        # Loop over beans
        for (bean.i in 1:n.beans) {
          bean.i.log <- data.i$bean.num == bean.i

          dv.i <- data.i[bean.i.log, dv.name]

          if (is.null(pointpars) == FALSE) {
            pointpars.bean.i <- pointpars.i[bean.i.log, ]
          }

          if (is.logical(dv.i)) {
            dv.i <- as.numeric(dv.i)
          }

          x.loc.i <- bean.mtx$x.loc[bean.i]

          # CALCULATE DENSITIES

          if (length(dv.i) > 3) { # only if n > 5

            dens.i <- density(dv.i, bw, adjust)

            dens.y.i <- dens.i$y
            dens.x.i <- dens.i$x

            # Rescale density according to width.max and width.min

            if (max(dens.y.i) < width.min) {
              dens.y.i <- dens.y.i / max(dens.y.i) * width.min
            }

            if (max(dens.y.i) > width.max) {
              dens.y.i <- dens.y.i / max(dens.y.i) * width.max
            }

            # adjust to cut.min and cut.max

            dens.x.plot.i <- dens.x.i
            dens.y.plot.i <- dens.y.i

            if (is.null(cut.min) == FALSE) {
              dens.x.plot.i <- dens.x.i[dens.x.i > cut.min]
              dens.y.plot.i <- dens.y.i[dens.x.i > cut.min]
            }


            if (is.null(cut.max) == FALSE) {
              dens.x.plot.i <- dens.x.i[dens.x.i < cut.max]
              dens.y.plot.i <- dens.y.i[dens.x.i < cut.max]
            }


            if (cap.beans == TRUE) {
              dens.x.plot.i <- dens.x.i[dens.x.i < max(dv.i) & dens.x.i > min(dv.i)]
              dens.y.plot.i <- dens.y.i[dens.x.i < max(dv.i) & dens.x.i > min(dv.i)]
            }
          }

          # BAR
          {
            rect(
              xleft = x.loc.i - width.max,
              ybottom = 0,
              xright = x.loc.i + width.max,
              ytop = summary$avg[bean.i],
              col = transparent(colors.df$bar.f.col[bean.i],
                trans.val = 1 - opac.df$bar.f.o[bean.i]
              ),
              border = transparent(colors.df$bar.b.col[bean.i],
                trans.val = 1 - opac.df$bar.b.o[bean.i]
              ),
              lwd = bar.b.lwd[bean.i]
            )
          }

          # BEAN
          {
            if (length(setdiff(dv.i, c(0, 1))) > 0 & length(dv.i) > 3) {
              polygon(
                c(
                  x.loc.i - dens.y.plot.i[1:(length(dens.x.plot.i))],
                  x.loc.i + rev(dens.y.plot.i[1:(length(dens.x.plot.i))])
                ),
                c(
                  dens.x.plot.i[1:(length(dens.x.plot.i))],
                  rev(dens.x.plot.i[1:(length(dens.x.plot.i))])
                ),
                col = transparent(colors.df$bean.f.col[bean.i],
                  trans.val = 1 - opac.df$bean.f.o[bean.i]
                ),
                border = transparent(colors.df$bean.b.col[bean.i],
                  trans.val = 1 - opac.df$bean.b.o[bean.i]
                ),
                lwd = bean.lwd[bean.i], lty = bean.lty[bean.i]
              )
            }
          }

          # POINTS
          {
            if (is.null(pointpars) == FALSE) {
              if ("jitter" %in% names(pointpars)) {
                my.jitter <- pointpars.i$jitter[bean.i.log]
              } else {
                my.jitter <- rnorm(length(dv.i), mean = 0, sd = jitter.val)
              }
            } else {
              my.jitter <- rnorm(length(dv.i), mean = 0, sd = jitter.val)
            }

            if (is.null(pointpars)) {
              # 1-color points
              if ((point.pch %in% 21:25) == FALSE) {
                points(
                  x = rep(x.loc.i, length(dv.i)) + my.jitter,
                  y = dv.i,
                  pch = point.pch,
                  col = transparent(colors.df$point.col[bean.i],
                    trans.val = 1 - opac.df$point.o[bean.i]
                  ),
                  cex = point.cex,
                  lwd = point.lwd
                )
              }

              # 2-color points
              if (point.pch %in% 21:25) {
                points(
                  x = rep(x.loc.i, length(dv.i)) + my.jitter,
                  y = dv.i,
                  pch = point.pch,
                  col = transparent(colors.df$point.col[bean.i],
                    trans.val = 1 - opac.df$point.o[bean.i]
                  ),
                  bg = transparent(colors.df$point.bg[bean.i],
                    trans.val = 1 - opac.df$point.o[bean.i]
                  ),
                  cex = point.cex,
                  lwd = point.lwd
                )
              }
            }

            if (is.null(pointpars) == FALSE) {
              if ("col" %in% names(pointpars)) {
                point.col <- transparent(pointpars$col[bean.i.log],
                  trans.val = 1 - opac.df$point.o[bean.i]
                )
              } else {
                point.col <- transparent(colors.df$point.col[bean.i],
                  trans.val = 1 - opac.df$point.o[bean.i]
                )
              }



              if ("bg" %in% names(pointpars)) {
                point.bg <- transparent(pointpars$bg[bean.i.log],
                  trans.val = 1 - opac.df$point.o[bean.i]
                )
              } else {
                point.bg <- transparent(colors.df$point.bg[bean.i],
                  trans.val = 1 - opac.df$point.o[bean.i]
                )
              }


              if ("pch" %in% names(pointpars)) {
                point.pch.i <- pointpars$pch[bean.i.log]
              } else {
                point.pch.i <- point.pch
              }


              if (point.pch %in% 21:25) {
                points(
                  x = rep(x.loc.i, length(dv.i)) + my.jitter,
                  y = dv.i,
                  pch = point.pch.i,
                  col = point.col,
                  bg = point.bg,
                  cex = point.cex,
                  lwd = point.lwd
                )
              } else {
                points(
                  x = rep(x.loc.i, length(dv.i)) + my.jitter,
                  y = dv.i,
                  pch = point.pch.i,
                  col = point.col,
                  cex = point.cex,
                  lwd = point.lwd
                )
              }


              if ("labels" %in% names(pointpars)) {
                if ("pos" %in% names(pointpars)) {
                  text.pos <- pointpars$pos[bean.i.log]
                } else {
                  text.pos <- NULL
                }

                if ("cex.labels" %in% names(pointpars)) {
                  cex.labels <- pointpars$cex[bean.i.log]
                } else {
                  cex.labels <- point.cex
                }

                if (is.null(text.pos) == FALSE) {
                  text(
                    x = rep(x.loc.i, length(dv.i)) + my.jitter,
                    y = dv.i,
                    labels = pointpars$labels[bean.i.log],
                    pos = text.pos,
                    cex = cex.labels
                  )
                } else {
                  text(
                    x = rep(x.loc.i, length(dv.i)) + my.jitter,
                    y = dv.i,
                    labels = pointpars$labels[bean.i.log],
                    cex = cex.labels
                  )
                }
              }
            }
          }


          # QUANTILES
          if (!is.null(quant)) {
            # set default line length if length is not given manually
            if (is.null(quant.length)) {
              quant.length <- c(rep(0.65, length(quant)))
            } else {
              quant.length <- rep(quant.length, length.out = length(quant))
            }
            if (is.null(quant.lwd)) {
              quant.lwd <- c(rep(0.75, length(quant)))
            } else {
              quant.lwd <- rep(quant.lwd, length.out = length(quant))
            }


            for (i in 1:length(quant)) {
              segments(x.loc.i + (quant.length[i] - width.max),
                quantile(dv.i, probs = quant[i]),
                x.loc.i - (quant.length[i] - width.max),
                quantile(dv.i, probs = quant[i]),
                col = colors.df$quant.col[bean.i],
                lwd = quant.lwd[i], lend = 3
              )
            }
            segments(x.loc.i, quantile(dv.i, probs = min(quant)),
              x.loc.i, quantile(dv.i, probs = max(quant)),
              col = colors.df$quant.col[bean.i], lwd = quant.lwd[1],
              lend = 3, lty = 1
            )
          }

          # INFERENCE BAND
          {
            if (length(dv.i) > 3 & sd(dv.i) > 0) {
              if (length(dv.i) <= 3) {
                message(paste("Note: Group ", bean.i, " had too few observations (", length(dv.i), ") for an inference band", sep = ""))
                message(paste("Note: Group", bean.i, "had no variance, so no inference band :("))
              }

              dens.inf.x <- dens.x.i[dens.x.i >= summary$inf.lb[bean.i] & dens.x.i <= summary$inf.ub[bean.i]]
              dens.inf.y <- dens.y.i[dens.x.i >= summary$inf.lb[bean.i] & dens.x.i <= summary$inf.ub[bean.i]]

              # Draw inf band

              if (inf.disp == "line") {
                segments(x.loc.i, summary$inf.lb[bean.i],
                  x.loc.i, summary$inf.ub[bean.i],
                  col = transparent(colors.df$inf.f.col[bean.i],
                    trans.val = 1 - opac.df$inf.f.o[bean.i]
                  ),
                  lwd = inf.lwd[bean.i]
                )

                # Add whiskers

                segments(rep(x.loc.i - width.max * .2, 2),
                  c(summary$inf.lb[bean.i], summary$inf.ub[bean.i]),
                  rep(x.loc.i + width.max * .2, 2),
                  c(summary$inf.lb[bean.i], summary$inf.ub[bean.i]),
                  col = transparent(colors.df$inf.f.col[bean.i],
                    trans.val = 1 - opac.df$inf.f.o[bean.i]
                  ),
                  lwd = inf.lwd[bean.i]
                )
              }


              if (inf.disp == "rect") {
                rect(x.loc.i - width.max * .7,
                  summary$inf.lb[bean.i],
                  x.loc.i + width.max * .7,
                  summary$inf.ub[bean.i],
                  col = transparent(colors.df$inf.f.col[bean.i],
                    trans.val = 1 - opac.df$inf.f.o[bean.i]
                  ),
                  lwd = inf.lwd[bean.i],
                  border = transparent(colors.df$inf.b.col[bean.i],
                    trans.val = 1 - opac.df$inf.b.o[bean.i]
                  )
                )
              }

              if (inf.disp == "bean") {
                polygon(
                  c(
                    x.loc.i - dens.inf.y[1:(length(dens.inf.x))],
                    x.loc.i + rev(dens.inf.y[1:(length(dens.inf.x))])
                  ),
                  c(
                    dens.inf.x[1:(length(dens.inf.x))],
                    rev(dens.inf.x[1:(length(dens.inf.x))])
                  ),
                  col = transparent(colors.df$inf.f.col[bean.i],
                    trans.val = 1 - opac.df$inf.f.o[bean.i]
                  ),
                  border = transparent(colors.df$inf.b.col[bean.i],
                    trans.val = 1 - opac.df$inf.b.o[bean.i]
                  ),
                  lwd = bean.lwd[bean.i]
                )
              }
            }
          }

          # AVERAGE LINE
          {
            if (inf.disp %in% c("line", "rect")) {
              segments(
                x0 = x.loc.i - width.max,
                y0 = summary$avg[bean.i],
                x1 = x.loc.i + width.max,
                y1 = avg.line.fun(dv.i),
                col = transparent(colors.df$avg.line.col[bean.i],
                  trans.val = 1 - opac.df$avg.line.o[bean.i]
                ),
                lwd = avg.line.lwd[bean.i],
                lend = 3
              )
            }

            if (inf.disp == "bean") {
              fun.loc <- which(abs(dens.x.i - avg.line.fun(dv.i)) == min(abs(dens.x.i - avg.line.fun(dv.i))))

              segments(x.loc.i - dens.y.i[fun.loc],
                summary$avg[bean.i],
                x.loc.i + dens.y.i[fun.loc],
                summary$avg[bean.i],
                col = transparent(colors.df$avg.line.col[bean.i],
                  trans.val = 1 - opac.df$avg.line.o[bean.i]
                ),
                lwd = avg.line.lwd[bean.i],
                lend = 3
              )
            }
          }
        }
      }

      # LABELS

      if (is.null(xaxt) == TRUE) {
        if (is.null(cex.lab)) {
          cex.lab <- 1 / ((n.subplots - 1) * .1 + 1)
        }


        if (is.null(xaxt)) {
          axis(
            side = 1,
            at = bean.loc,
            labels = rep("", n.beans),
            lwd = 0,
            lwd.ticks = 1,
            cex.axis = cex.axis
          )
        }

        # IV 1 Labels

        mtext(bean.mtx[, 1],
          side = 1,
          at = bean.mtx$x.loc,
          line = 1,
          cex = cex.names
        )


        # IV 2 labels

        if (beside == TRUE & n.iv > 1) {
          mtext(iv.names[2], side = 1, line = 2.5, at = par("usr")[1], adj = 1, cex = cex.names)

          mtext(iv.names[1], side = 1, line = 1, at = par("usr")[1], adj = 1, cex = cex.names)

          text.loc <- (iv.lengths[1] + 1) / 2 * (2 * (1:iv.lengths[2]) - 1)

          mtext(
            text = unique(bean.mtx[, 2]),
            side = 1,
            line = 2.5,
            at = text.loc,
            cex = cex.lab
          )
        }

        if (beside == FALSE & n.iv > 1) {
          # mtext(iv.names[2], side = 1, line = 2.5, at = par("usr")[1], adj = 1, cex = cex.names)

          text.loc <- (iv.lengths[1] + 1) / 2 * (2 * (1:iv.lengths[2]) - 1)

          mtext(iv.names[1], side = 1, line = 2.5, at = text.loc, cex = cex.names)


          mtext(
            text = paste(iv.names[2], "=", unique(bean.mtx[, 2])),
            side = 3,
            line = 1,
            at = text.loc,
            cex = cex.lab
          )
        }
      }
    }


    summary.ls[[subplot.i]] <- summary
  }

  summary.df <- do.call(summary.ls, what = "rbind")
  summary.df <- summary.df[, c(all.iv.names, setdiff(names(summary.df), all.iv.names))]

  output.ls <- list(
    "summary" = summary.df,
    "avg.line.fun" = deparse(substitute(avg.line.fun)),
    "inf.method" = inf.method,
    "inf.p" = inf.p
  )

  ## Reset plotting space
  par(mfrow = c(1, 1))


  if (plot == FALSE) {
    return(output.ls)
  }
}