R/visualization.R

In DataScienceSalon/movies: Movie Rating Model and Predictor

#==============================================================================#
#                             Visualization Functions                          #
#==============================================================================#

#------------------------------------------------------------------------------#
#                     Frequency and Proportion Bar Plot                        #
#------------------------------------------------------------------------------#
#' plotFreqProp
#'
#' \code{plotFreqPropTbl} Renders a stacked bar plot and a contingency table
#' showing frequencies and proportions
#'
#' @param data Data frame or vector containing a single categorical factor variable
#' @param yLab Capital case character string describing the y variable
#' @param xLab Capital case character string containing the name of the variable x variable
#' @param title Capital case character string for the title of the plot
#' @param order Character c('a', 'd') indicating whether to order by frequency ascending (a), or descending (d)
#'
#' @return List containing a contingency table and a stacked bar plot.
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotFreqProp <- function(data, yLab = "Movies", xLab, title = NULL, order = NULL) {

  # Format title
  if (is.null(title)) {
    title <- paste(yLab, "by", xLab)
  }

  # Format Data
  df <- as.data.frame(data)
  df <- df %>%  group_by(.[[1]]) %>%
    summarize(N = n()) %>%
    mutate(`Min N` = getPSampleSize(N / sum(N)),
           Proportion = round(N / sum(N), 2),
           Cumulative = round(cumsum(Proportion), 2),
           pos = N / 2)
  colnames(df)[1] <- xLab

  # Order data
  if (!is.null(order)) {
    if (order ==  "a") {
      df <- df %>% arrange(N)
      } else if (order == "d") {
        df <- df %>% arrange(desc(N))
      }
  }

  # Render plot
  myPal <- colorRampPalette(RColorBrewer::brewer.pal(11, "PiYG"))
  barPlot <- ggplot2::ggplot(data = df,
                             ggplot2::aes(x = df[[1]],
                                          y = df[[2]],
                                          fill = df[[1]]))  +
    ggplot2::geom_bar(stat='identity') +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::geom_text(
      data = df,
      ggplot2::aes(x = df[[1]],
                   y = df[[6]],
                   label = paste0(df[[2]], " (",
                                  round(df[[4]] * 100, 0),
                                  "%)")),
      colour="black", family="Tahoma", size = 8) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans"),
                   axis.title.x = ggplot2::element_blank(),
                   axis.ticks.x = ggplot2::element_blank(),
                   axis.text.x = ggplot2::element_blank(),
                   legend.position = "right") +
    ggplot2::scale_fill_manual(values = myPal(length(df[[1]]))) +
    ggplot2::ggtitle(title) +
    ggplot2::ylab(yLab) +
    ggplot2::labs(fill = xLab)

    visual <- list(
      stats = df[,1:5],
      plot = barPlot
    )
    return(visual)
}


#------------------------------------------------------------------------------#
#                               Plot Histogram                                 #
#------------------------------------------------------------------------------#
#' plotHist
#'
#' \code{plotHist} Renders a histogram with a normal curve as well as a table
#' of summary statistics.
#'
#' @param data Data frame containing a single numeric variable
#' @param xLab Capital case character string the group or subset of data being printed (optional)
#' @param yLab Capital case character string describing y is being printed
#' @param title Capital case character string for the title of the plot
#'
#' @return List containing a summary statistics and a histogram
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotHist <- function(data, xLab = NULL, yLab, title = NULL) {

  if (is.null(title)) {
    if (is.null(xLab)) {
      title <- yLab
    } else {
      title <- paste(yLab, "by", xLab)
    }
  }

  hist <- ggplot2::ggplot(data = data,
                        ggplot2::aes(x = data[[1]])) +
    ggplot2::geom_histogram(position ='identity', color = "white", fill = "palegreen4") +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans"),
                   axis.title.x = ggplot2::element_blank(),
                   axis.ticks.x = ggplot2::element_blank()) +
    ggplot2::ggtitle(title) +
    ggplot2::ylab(yLab)

  return(hist)
}

#------------------------------------------------------------------------------#
#                               Plot Quantile                                  #
#------------------------------------------------------------------------------#
#' plotQQ
#'
#' \code{plotQQ} Renders a histogram with a normal curve as well as a table
#' of summary statistics.
#'
#' @param data Data frame or vector containing a single numeric variable
#' @param xLab Capital case character string containing the name of the grouping or subset variable (optional)
#' @param yLab Capital case character string containing the name of the y variable
#' @param title Capital case character string for the title of the plot
#'
#' @return List containing a summary statistics and QQ plot
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotQQ <- function(data, xLab = NULL, yLab, title = NULL) {

  if (is.null(title)) {
    if (is.null(xLab)) {
      title <- paste("Normal Q-Q Plot:", yLab)
    } else {
      title <- paste("Normal Q-Q Plot:", yLab, "by", xLab)
    }
  }


  # Render QQ Plot
  qq <- ggplot2::ggplot(data = data, mapping = aes(sample = data[[1]])) +
    qqplotr::stat_qq_band() +
    qqplotr::stat_qq_line() +
    qqplotr::stat_qq_point() +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans")) +
    ggplot2::labs(x = "Theoretical Quantiles", y = "Sample Quantiles") +
    ggplot2::ggtitle(title)

  return(qq)
}

#------------------------------------------------------------------------------#
#                                 Boxplot                                      #
#------------------------------------------------------------------------------#
#' plotBox
#'
#' \code{plotBox} Renders a single or grouped box plot.
#'
#' @param data Data frame or vector containing two columns:
#'          1: Categorical independent variable (x)
#'          2: Numeric response variable (y)
#' @param xLab Capital case character string containing the name of the x variable (optional)
#' @param yLab Capital case character string containing the name of the y variable
#' @param title Character case character string containing the title for the plot
#'
#' @return List containing a contingency table and a stacked bar plot.
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotBox <- function(data, xLab = NULL, yLab, title = NULL) {

  if (length(data) > 2) stop(paste("Error in plotBox: Dimension of data frame must be 1 or 2, not", length(data)))
  if (length(data) == 1) {
    data <- data.frame(x = rep("x", nrow(data)),
                       y = data[[1]], row.names = NULL)
  } else {
    data <- data.frame(x = data[[1]],
                       y = data[[2]],
                       row.names = NULL)
  }

  # Format title
  if (is.null(title)) {
   if (is.null(xLab)) {
     title <- yLab
   } else {
     title <- paste(yLab, "by", xLab)
   }
  }

  # Render plot
  myPal <- colorRampPalette(RColorBrewer::brewer.pal(11, "PiYG"))

  bp <- ggplot2::ggplot(data = data,
                             ggplot2::aes(x = data$x,
                                          y = data$y,
                                          fill = data$x))  +
    ggplot2::geom_boxplot(outlier.colour = "black") +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans"),
                   axis.title.x = ggplot2::element_blank(),
                   axis.ticks.x = ggplot2::element_blank(),
                   axis.text.x = ggplot2::element_blank(),
                   legend.position = "right") +
    ggplot2::ggtitle(title) +
    ggplot2::scale_fill_manual(values = myPal(length(unique(data$x))))

  if (is.null(xLab)) {
    bp <- bp + ggplot2::coord_flip() +
      ggplot2::theme(legend.position = "none",
                     axis.text.y = ggplot2::element_blank(),
                     axis.title.y = ggplot2::element_blank())
  } else {
    bp <- bp + ggplot2::labs(y = yLab, fill = xLab)
  }

  return(bp)
}


#------------------------------------------------------------------------------#
#                               Plot Scatterplot                               #
#------------------------------------------------------------------------------#
#' plotScatter
#'
#' \code{plotScatter} Renders a scatterplot for two numerical variablesa
#'
#' @param data Data frame containing the quantitative variables
#' @param xLab Capital case character string containing the name of the grouping or subset variable
#' @param yLab Capital case character string containing the name of the y variable
#' @param title Capital case character string for title of plot
#'
#' @return Scatterplot object
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotScatter <- function(data, xLab, yLab, title = NULL) {

  if (is.null(title)) {
    title <- paste(yLab, "by", xLab)
  }

  myPal <- colorRampPalette(RColorBrewer::brewer.pal(11, "PiYG"))

  scatter <- ggplot2::ggplot(data = data,
                             ggplot2::aes(x = as.numeric(unlist(data[,1])),
                                          y = as.numeric(unlist(data[,2])))) +
    ggplot2::geom_point() +
    ggplot2::geom_smooth() +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::geom_smooth(method = lm, se = FALSE) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans"),
                   legend.position = "right") +
    ggplot2::labs(x = xLab, y = yLab) +
    ggplot2::ggtitle(title)

  return(scatter)

}

#------------------------------------------------------------------------------#
#                             Plot Residual QQ                                 #
#------------------------------------------------------------------------------#
#' plotResQQ
#'
#' \code{plotResQQ} Plots regression diagnostics
#'
#' @param mod Linear model object
#' @param yLab Capital case character string for response variable
#'
#' @return resQQ Residual Normal QQ Plot
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotResQQ <- function(mod, yLab) {

  # Obtain diagnostics and render plot
  resQQ <- gg_diagnose(mod, theme = theme_minimal(), plot.all = FALSE)
  resQQ <- resQQ$qqplot +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans")) +
    ggplot2::ggtitle(paste("Normal-QQ Plot:", yLab))

  return(resQQ)
}

#------------------------------------------------------------------------------#
#                           Plot Residual vs. Fit                              #
#------------------------------------------------------------------------------#
#' plotResFit
#'
#' \code{plotResFit} Plots regression diagnostics
#'
#' @param mod Linear model object
#' @param yLab Capital case character string for response variable
#'
#' @return resFit Residual vs. Fit Plot
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotResFit <- function(mod, yLab) {

  # Obtain diagnostics and render plot
  resFit <- gg_diagnose(mod, theme = theme_minimal(), plot.all = FALSE)
  resFit <- resFit$res_fitted +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans")) +
    ggplot2::ggtitle(paste("Residuals vs. Fit:", yLab))

  return(resFit)
}

#------------------------------------------------------------------------------#
#                           Plot Residual vs. Leverage                         #
#------------------------------------------------------------------------------#
#' plotResLeverage
#'
#' \code{plotResLeverage} Plots regression diagnostics
#'
#' @param mod Linear model object
#' @param yLab Capital case character string for response variable
#'
#' @return resLeverage Residual vs. Leverage Plot
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotResLeverage <- function(mod, yLab) {

  # Obtain diagnostics and render plot
  resLeverage <- gg_diagnose(mod, theme = theme_minimal(), plot.all = FALSE)
  resLeverage <- resLeverage$resleverage +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans")) +
    ggplot2::ggtitle(paste("Residuals vs. Leverage:", yLab))

  return(resLeverage)
}

#------------------------------------------------------------------------------#
#                           Plot Residuals Cooks Distance                      #
#------------------------------------------------------------------------------#
#' plotCooks
#'
#' \code{plotCooks} Plots cooks distance
#'
#' @param mod Linear model object
#' @param yLab Capital case character string for response variable
#'
#' @return cooks Cooks distance plot for residuals
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotCooks <- function(mod, yLab) {

  # Obtain diagnostics and render plot
  cooks <- gg_diagnose(mod, theme = theme_minimal(), plot.all = FALSE)
  cooks <- cooks$cooksd +
    ggplot2::theme_minimal(base_size = 24) +
    ggplot2::theme(text = ggplot2::element_text(family="Open Sans")) +
    ggplot2::ggtitle(paste("Cooks Distance:", yLab))

  return(cooks)
}

#------------------------------------------------------------------------------#
#                     Plot Residuals for all Predictors                        #
#------------------------------------------------------------------------------#
#' plotResAll
#'
#' \code{plotResAll} Plots cooks distance
#'
#' @param mod Linear model object
#'
#' @return list containing ressidual v predictor plots
#' @author John James, \email{jjames@@datasciencesalon.org}
#' @family visualization functions
#' @export
plotResAll <- function(mod) {

  # Obtain diagnostics and render plot
  p <- gg_resX(mod, plot.all = FALSE)

  resAll <- lapply(p, function(x) {
    df <- data.frame(x = p[[1]][[1]][[2]],
                     y = p[[1]][[1]][[1]])
    title <- p[[1]][[9]][[1]]
    xLab <- p[[1]][[9]][[3]]
    yLab <- p[[1]][[9]][[4]]
    plotBox(data = df, xLab = xLab, yLab = yLab, title = title)
  })

  return(resAll)
}