7

我想charts.PerformanceSummaryPerformanceAnalytics包中提供基本功能的“ggplot 版本”,因为我认为 ggplot 通常更漂亮,理论上在编辑图像方面更强大。我已经相当接近了,但有一些问题需要一些帮助。即:

  1. 减少图例占用的空间量,当上面有超过 10 行时,它会变得可怕/丑陋......(只需线条颜色和名称就足够了)
  2. 增加 Daily_Returns 方面的大小以匹配图表的大小。PerformanceSummary 在PerformanceAnalytics
  3. 有一个选项可以指定在 Daily_Returns 方面的每日回报系列中显示哪些资产,而不是总是使用第一列,这比发生在charts.PerformanceSummary

如果有更好的方法可以使用gridExtra而不是方面来做到这一点......我并不反对人们向我展示这看起来会更好......

这里的问题是美学,我想可能很容易操作,因为 PerformanceAnalytics 已经有一个很好的工作示例,我只是想让它更漂亮/更专业......

除了奖励积分之外,我还希望能够在每个资产的图表上方或下方或侧面的某处显示一些与之相关的性能统计数据......不太确定在哪里最好显示或显示此信息。

此外,如果有人对此提出建议,我并不反对那些建议清理我的代码的部分。

这是我的可重现示例...

首先生成返回数据:

require(xts)
X.stock.rtns <- xts(rnorm(1000,0.00001,0.0003), Sys.Date()-(1000:1))
Y.stock.rtns <- xts(rnorm(1000,0.00003,0.0004), Sys.Date()-(1000:1))
Z.stock.rtns <- xts(rnorm(1000,0.00005,0.0005), Sys.Date()-(1000:1))
rtn.obj <- merge(X.stock.rtns , Y.stock.rtns, Z.stock.rtns)
colnames(rtn.obj) <- c("x.stock.rtns","y.stock.rtns","z.stock.rtns")

我想从以下结果中复制图像:

require(PerformanceAnalytics)
charts.PerformanceSummary(rtn.obj, geometric=TRUE)

目标

这是我迄今为止的尝试......

gg.charts.PerformanceSummary <- function(rtn.obj, geometric=TRUE, main="",plot=TRUE){

    # load libraries
suppressPackageStartupMessages(require(ggplot2))
suppressPackageStartupMessages(require(scales))
suppressPackageStartupMessages(require(reshape))
suppressPackageStartupMessages(require(PerformanceAnalytics))
    # create function to clean returns if having NAs in data
    clean.rtn.xts <- function(univ.rtn.xts.obj,na.replace=0){
    univ.rtn.xts.obj[is.na(univ.rtn.xts.obj)]<- na.replace
    univ.rtn.xts.obj
}
    # Create cumulative return function
cum.rtn <- function(clean.xts.obj, g=TRUE){
    x <- clean.xts.obj
    if(g==TRUE){y <- cumprod(x+1)-1} else {y <- cumsum(x)}
    y
}
    # Create function to calculate drawdowns
dd.xts <- function(clean.xts.obj, g=TRUE){
    x <- clean.xts.obj
    if(g==TRUE){y <- Drawdowns(x)} else {y <- Drawdowns(x,geometric=FALSE)}
    y
}
    # create a function to create a dataframe to be usable in ggplot to replicate charts.PerformanceSummary
cps.df <- function(xts.obj,geometric){
    x <- clean.rtn.xts(xts.obj)
    series.name <- colnames(xts.obj)[1]
    tmp <- cum.rtn(x,geometric)
    tmp$rtn <- x
    tmp$dd <- dd.xts(x,geometric)
    colnames(tmp) <- c("Cumulative_Return","Daily_Return","Drawdown")
    tmp.df <- as.data.frame(coredata(tmp))
    tmp.df$Date <- as.POSIXct(index(tmp))
    tmp.df.long <- melt(tmp.df,id.var="Date")
    tmp.df.long$asset <- rep(series.name,nrow(tmp.df.long))
    tmp.df.long
}
# A conditional statement altering the plot according to the number of assets
if(ncol(rtn.obj)==1){
            # using the cps.df function
    df <- cps.df(rtn.obj,geometric)
            # adding in a title string if need be
    if(main==""){
        title.string <- paste0(df$asset[1]," Performance")
    } else {
        title.string <- main
    }
            # generating the ggplot output with all the added extras....
    gg.xts <- ggplot(df, aes_string(x="Date",y="value",group="variable"))+
                facet_grid(variable ~ ., scales="free", space="free")+
                geom_line(data=subset(df,variable=="Cumulative_Return"))+
                geom_bar(data=subset(df,variable=="Daily_Return"),stat="identity")+
                geom_line(data=subset(df,variable=="Drawdown"))+
                ylab("")+
                geom_abline(intercept=0,slope=0,alpha=0.3)+
                ggtitle(title.string)+
                theme(axis.text.x = element_text(angle = 45, hjust = 1))+
                scale_x_datetime(breaks = date_breaks("6 months"), labels = date_format("%d/%m/%Y"))

} else {
            # a few extra bits to deal with the added rtn columns
    no.of.assets <- ncol(rtn.obj)
    asset.names <- colnames(rtn.obj)
    df <- do.call(rbind,lapply(1:no.of.assets, function(x){cps.df(rtn.obj[,x],geometric)}))
    df$asset <- ordered(df$asset, levels=asset.names)
    if(main==""){
        title.string <- paste0(df$asset[1]," Performance")
    } else {
        title.string <- main
    }
    if(no.of.assets>5){legend.rows <- 5} else {legend.rows <- no.of.assets}
    gg.xts <- ggplot(df, aes_string(x="Date", y="value",group="asset"))+
      facet_grid(variable~.,scales="free",space="free")+
      geom_line(data=subset(df,variable=="Cumulative_Return"),aes(colour=factor(asset)))+
      geom_bar(data=subset(df,variable=="Daily_Return"),stat="identity",aes(fill=factor(asset),colour=factor(asset)),position="dodge")+
      geom_line(data=subset(df,variable=="Drawdown"),aes(colour=factor(asset)))+
      ylab("")+
      geom_abline(intercept=0,slope=0,alpha=0.3)+
      ggtitle(title.string)+
      theme(legend.title=element_blank(), legend.position=c(0,1), legend.justification=c(0,1),
            axis.text.x = element_text(angle = 45, hjust = 1))+
      guides(col=guide_legend(nrow=legend.rows))+
      scale_x_datetime(breaks = date_breaks("6 months"), labels = date_format("%d/%m/%Y"))

}

assign("gg.xts", gg.xts,envir=.GlobalEnv)
if(plot==TRUE){
    plot(gg.xts)
} else {}

}
# seeing the ggplot equivalent....
gg.charts.PerformanceSummary(rtn.obj, geometric=TRUE)

结果

4

2 回答 2

14

我就是在找那个。你已经很接近了。站在你的肩膀上,我能够解决一些问题。

编辑(2015 年 5 月 9 日):Drawdown()现在可以通过三冒号运算符调用该函数, PerformanceAnalytics:::Drawdown(). 编辑下面的代码以反映此更改。编辑(2018 年 4 月 22 日): show_guide已弃用并由show.legend.

require(xts)

X.stock.rtns <- xts(rnorm(1000,0.00001,0.0003), Sys.Date()-(1000:1))
Y.stock.rtns <- xts(rnorm(1000,0.00003,0.0004), Sys.Date()-(1000:1))
Z.stock.rtns <- xts(rnorm(1000,0.00005,0.0005), Sys.Date()-(1000:1))
rtn.obj <- merge(X.stock.rtns , Y.stock.rtns, Z.stock.rtns)
colnames(rtn.obj) <- c("x","y","z")

# advanced charts.PerforanceSummary based on ggplot
gg.charts.PerformanceSummary <- function(rtn.obj, geometric = TRUE, main = "", plot = TRUE)
{

    # load libraries
    suppressPackageStartupMessages(require(ggplot2))
    suppressPackageStartupMessages(require(scales))
    suppressPackageStartupMessages(require(reshape))
    suppressPackageStartupMessages(require(PerformanceAnalytics))

    # create function to clean returns if having NAs in data
    clean.rtn.xts <- function(univ.rtn.xts.obj,na.replace=0){
        univ.rtn.xts.obj[is.na(univ.rtn.xts.obj)]<- na.replace
        univ.rtn.xts.obj  
    }

    # Create cumulative return function
    cum.rtn <- function(clean.xts.obj, g = TRUE)
    {
        x <- clean.xts.obj
        if(g == TRUE){y <- cumprod(x+1)-1} else {y <- cumsum(x)}
        y
    }

    # Create function to calculate drawdowns
    dd.xts <- function(clean.xts.obj, g = TRUE)
    {
        x <- clean.xts.obj
        if(g == TRUE){y <- PerformanceAnalytics:::Drawdowns(x)} else {y <- PerformanceAnalytics:::Drawdowns(x,geometric = FALSE)}
        y
    }

    # create a function to create a dataframe to be usable in ggplot to replicate charts.PerformanceSummary
    cps.df <- function(xts.obj,geometric)
    {
        x <- clean.rtn.xts(xts.obj)
        series.name <- colnames(xts.obj)[1]
        tmp <- cum.rtn(x,geometric)
        tmp$rtn <- x
        tmp$dd <- dd.xts(x,geometric)
        colnames(tmp) <- c("Index","Return","Drawdown") # names with space
        tmp.df <- as.data.frame(coredata(tmp))
        tmp.df$Date <- as.POSIXct(index(tmp))
        tmp.df.long <- melt(tmp.df,id.var="Date")
        tmp.df.long$asset <- rep(series.name,nrow(tmp.df.long))
        tmp.df.long
    }

    # A conditional statement altering the plot according to the number of assets
    if(ncol(rtn.obj)==1)
    {
        # using the cps.df function
        df <- cps.df(rtn.obj,geometric)
        # adding in a title string if need be
        if(main == ""){
            title.string <- paste("Asset Performance")
        } else {
            title.string <- main
        }
    
        gg.xts <- ggplot(df, aes_string( x = "Date", y = "value", group = "variable" )) +
            facet_grid(variable ~ ., scales = "free_y", space = "fixed") +
            geom_line(data = subset(df, variable == "Index")) +
            geom_bar(data = subset(df, variable == "Return"), stat = "identity") +
            geom_line(data = subset(df, variable == "Drawdown")) +
            geom_hline(yintercept = 0, size = 0.5, colour = "black") +
            ggtitle(title.string) +
            theme(axis.text.x = element_text(angle = 0, hjust = 1)) +
            scale_x_datetime(breaks = date_breaks("6 months"), labels = date_format("%m/%Y")) +
            ylab("") +
            xlab("")
    
    } 
    else 
    {
        # a few extra bits to deal with the added rtn columns
        no.of.assets <- ncol(rtn.obj)
        asset.names <- colnames(rtn.obj)
        df <- do.call(rbind,lapply(1:no.of.assets, function(x){cps.df(rtn.obj[,x],geometric)}))
        df$asset <- ordered(df$asset, levels=asset.names)
        if(main == ""){
            title.string <- paste("Asset",asset.names[1],asset.names[2],asset.names[3],"Performance")
        } else {
            title.string <- main
        }
    
        if(no.of.assets>5){legend.rows <- 5} else {legend.rows <- no.of.assets}
    
        gg.xts <- ggplot(df, aes_string(x = "Date", y = "value" )) +
        
            # panel layout
            facet_grid(variable~., scales = "free_y", space = "fixed", shrink = TRUE, drop = TRUE, margin = 
                           , labeller = label_value) + # label_value is default
        
            # display points for Index and Drawdown, but not for Return
            geom_point(data = subset(df, variable == c("Index","Drawdown"))
                       , aes(colour = factor(asset), shape = factor(asset)), size = 1.2, show.legend = TRUE) + 
        
            # manually select shape of geom_point
            scale_shape_manual(values = c(1,2,3)) + 
        
            # line colours for the Index
            geom_line(data = subset(df, variable == "Index"), aes(colour = factor(asset)), show.legend = FALSE) +
        
            # bar colours for the Return
            geom_bar(data = subset(df,variable == "Return"), stat = "identity"
                     , aes(fill = factor(asset), colour = factor(asset)), position = "dodge", show.legend = FALSE) +
        
            # line colours for the Drawdown
            geom_line(data = subset(df, variable == "Drawdown"), aes(colour = factor(asset)), show.legend = FALSE) +
        
            # horizontal line to indicate zero values
            geom_hline(yintercept = 0, size = 0.5, colour = "black") +
        
            # horizontal ticks
            scale_x_datetime(breaks = date_breaks("6 months"), labels = date_format("%m/%Y")) +
        
            # main y-axis title
            ylab("") +
        
            # main x-axis title
            xlab("") +
        
            # main chart title
            ggtitle(title.string)
    
        # legend 
    
        gglegend <- guide_legend(override.aes = list(size = 3))
    
        gg.xts <- gg.xts + guides(colour = gglegend, size = "none") +
        
            # gglegend <- guide_legend(override.aes = list(size = 3), direction = "horizontal") # direction overwritten by legend.box?
            # gg.xts <- gg.xts + guides(colour = gglegend, size = "none", shape = gglegend) + # Warning: "Duplicated override.aes is ignored"
        
            theme( legend.title = element_blank()
                   , legend.position = c(0,1)
                   , legend.justification = c(0,1)
                   , legend.background = element_rect(colour = 'grey')
                   , legend.key = element_rect(fill = "white", colour = "white")
                   , axis.text.x = element_text(angle = 0, hjust = 1)
                   , strip.background = element_rect(fill = "white")
                   , panel.background = element_rect(fill = "white", colour = "white")
                   , panel.grid.major = element_line(colour = "grey", size = 0.5) 
                   , panel.grid.minor = element_line(colour = NA, size = 0.0)
            )
    
    }

    assign("gg.xts", gg.xts,envir=.GlobalEnv)
    if(plot == TRUE){
        plot(gg.xts)
    } else {}

}

# display chart
gg.charts.PerformanceSummary(rtn.obj, geometric = TRUE)

面板大小的控制在 facet_grid 内部:facet_grid(variable ~ ., scales = "free_y", space = "fixed")。这些选项的作用在手册中进行了解释,引用:

scales:尺度是否在所有方面共享(默认值,“固定”),或者它们是否在行(“free_x”)、列(“free_y”)或行和列(“free”)之间变化

space:如果“固定”,默认情况下,所有面板都具有相同的大小。如果“free_y”,它们的高度将与 y 刻度的长度成正比;如果“free_x”,它们的宽度将与 x 刻度的长度成正比;或者如果“免费”,高度和宽度都会有所不同。除非适当的比例也发生变化,否则此设置无效。

更新:标签

自定义标签可以通过以下函数获得:

# create a function to store fancy axis labels 

    my_labeller <- function(var, value){ # from the R Cookbook
        value <- as.character(value)
        if (var=="variable") 
        {
              value[value=="Index"] <- "Cumulative Returns"
              value[value=="Return"] <- "Daily Returns"
              value[value=="Drawdown"] <- "Drawdown"
        }
        return(value)
    }

并将 labeller 选项设置为“labeller = my_labeller”

更新:背景

背景、网格线、颜色等的外观可以在 theme() 函数中进行控制:上面的代码已经更新以反映这些变化。

在此处输入图像描述

于 2013-03-30T07:24:33.903 回答
1

有关图例的大小,请参阅 ?theme。图例的大部分方面都可以通过那里进行调整...我想您要调整的是 legend.key.size 以及 legend.background 以删除每个图例周围的框...

刻面中每个面板的大小有点复杂。我有一个技巧,可让您在调用 facet_grid 时指定每个面板的相对大小,但它需要从源等安装...更好的解决方案是将您的绘图转换为 gtable 对象并修改它...假设您的绘图称为 p:

require(gtable)
require(grid)

pTable <- ggplot_gtable(ggplot_build(p))
pTable$heights[[4]] <- unit(2, 'null')

grid.newpage()
grid.draw(pTable)

这将使顶部面板的高度成为其他面板大小的两倍......它是 pTable$heights[[4]] 而不是 pTable$heights[[1]] 的原因是分面面板不是情节中的顶级grobs。

我将避免比这更具体,因为您最好自己探索 gtable 的属性(而且因为我没有时间)

最好的

托马斯

于 2013-02-15T10:17:04.913 回答