引言
在进行数据分析时,免不了对结果进行可视化。那么,什么样的图形才最适合自己的数据呢?一个有效的图形应具备以下特点:
- 能正确传递信息,而不会产生歧义;
- 样式简单,但是易于理解;
- 添加的图形美学应辅助理解信息;
- 图形上不应出现冗余无用的信息。
本系列推文,小编将汇总可视化中常用 7 大类型图形,供读者参考。
每类制作成一篇推文,主要参考资料为:Top 50 ggplot2 Visualizations[1]。其他类似功能网站,资料包括:
- 庄闪闪的可视化笔记——常用图形[2]
- R Graph Gallery[3]
- 《R 语言教程》——ggplot 的各种图形[4]
系列目录
本文主要介绍第六部分:变化趋势图形。
加载数据集
使用 ggplot2
包中自带数据集作为示例数据集。
library(ggplot2) library(plotrix) data("midwest", package = "ggplot2") #加载数据集
midwest 数据集
全局主题设置
全局配色、主题设置。注意,本文使用离散色阶,如果需要使用连续色阶,则需要重写。
options(scipen=999) # 关掉像 1e+48 这样的科学符号 # 颜色设置(灰色系列) cbp1 <- c("#999999", "#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2", "#D55E00", "#CC79A7") # 颜色设置(黑色系列) cbp2 <- c("#000000", "#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2", "#D55E00", "#CC79A7") ggplot <- function(...) ggplot2::ggplot(...) + scale_color_manual(values = cbp1) + scale_fill_manual(values = cbp1) + # 注意: 使用连续色阶时需要重写 theme_bw()
6 变化趋势
6.1 时间序列图:基于时间序列对象(ts)
ggfortify
包中的 autoplot()
可以对时间序列直接绘图。在此基础上,添加其他 ggplot 相关函数。
library(ggplot2) library(ggfortify) theme_set(theme_classic()) # 绘图 autoplot(AirPassengers) + labs(title="AirPassengers") + theme(plot.title = element_text(hjust=0.5))
时间序列图:基于时间序列对象(ts)
6.2 时间序列图:基于数据框
library(ggplot2) theme_set(theme_classic()) # 使用默认的时间跨度 ggplot(economics, aes(x=date)) + geom_line(aes(y=pce)) + labs(title="Time Series Chart", caption="Source: Economics")
economics 数据集预览:来自 ggplot2 包
时间序列图:基于数据框
如果想设置特定的时间间隔,则需要使用 scale_x_date()
函数。
library(ggplot2) library(lubridate) theme_set(theme_bw()) economics_m <- economics[1:24, ] # 设定时间跨度为一个月 lbls <- paste0(month.abb[month(economics_m$date)], " ", lubridate::year(economics_m$date)) brks <- economics_m$date # 绘图 ggplot(economics_m, aes(x=date)) + geom_line(aes(y=pce)) + scale_x_date(labels = lbls, breaks = brks) + # change to monthly ticks and labels theme(axis.text.x = element_text(angle = 90, vjust=0.5), # rotate x axis text panel.grid.minor = element_blank()) # turn off minor grid
时间序列图:跨度为一个月
设置时间跨度为 1 年:
library(ggplot2) library(lubridate) theme_set(theme_bw()) economics_y <- economics[1:90, ] # 选择一年数据 # X轴文本的标签和 break 值 brks <- economics_y$date[seq(1, length(economics_y$date), 12)] lbls <- lubridate::year(brks) # 绘图 ggplot(economics_y, aes(x=date)) + geom_line(aes(y=pce)) + scale_x_date(labels = lbls, breaks = brks) + # change to monthly ticks and labels theme(axis.text.x = element_text(angle = 90, vjust=0.5), # rotate x axis text panel.grid.minor = element_blank())
时间序列图:时间跨度为 1 年
6.3 多个时间序列
在本例中,基于长数据格式进行可视化。这意味着,所有列的列名和各自的值被存放在两个变量中(分别是 variable
和 value
)。
data(economics_long, package = "ggplot2") head(economics_long)
在下面的代码中,在 geom_line()
函数中设置绘图对象为 value
,颜色匹配对象为 variable
。这样,只要调用一次 geom_line,就会绘制多条彩色线,每条线代表 variable
列中的每个唯一 value
。scale_x_date()
将更改 X 轴断点和标签,scale_color_manual
将更改行颜色。
library(ggplot2) library(lubridate) theme_set(theme_bw()) df <- economics_long[economics_long$variable %in% c("psavert", "uempmed"), ] df <- df[lubridate::year(df$date) %in% c(1967:1981), ] # labels and breaks for X axis text brks <- df$date[seq(1, length(df$date), 12)] lbls <- lubridate::year(brks) # 绘图 ggplot(df, aes(x=date)) + geom_line(aes(y=value, col=variable)) + labs(title="Time Series of Returns Percentage", subtitle="Drawn from Long Data format", caption="Source: Economics", y="Returns %", color=NULL) + # title and caption scale_x_date(labels = lbls, breaks = brks) + # change to monthly ticks and labels scale_color_manual(labels = c("psavert", "uempmed"), values = c("psavert"="#00ba38", "uempmed"="#f8766d")) + # line color theme(axis.text.x = element_text(angle = 90, vjust=0.5, size = 8), # rotate x axis text panel.grid.minor = element_blank()) # turn off minor grid
多个时间序列
如果从一个宽格式创建一个时间序列,则必须通过对每条线调用一次 geom_line()
。因此,默认情况下不会绘制图例,需要手动添加。
library(ggplot2) library(lubridate) theme_set(theme_bw()) df <- economics[, c("date", "psavert", "uempmed")] df <- df[lubridate::year(df$date) %in% c(1967:1981), ] # labels and breaks for X axis text brks <- df$date[seq(1, length(df$date), 12)] lbls <- lubridate::year(brks) # plot ggplot(df, aes(x=date)) + geom_line(aes(y=psavert, col="psavert")) + geom_line(aes(y=uempmed, col="uempmed")) + labs(title="Time Series of Returns Percentage", subtitle="Drawn From Wide Data format", caption="Source: Economics", y="Returns %") + # title and caption scale_x_date(labels = lbls, breaks = brks) + # change to monthly ticks and labels scale_color_manual(name="", values = c("psavert"="#00ba38", "uempmed"="#f8766d")) + # line color theme(panel.grid.minor = element_blank()) # turn off minor grid
多个时间序列
6.4 堆叠面积图
堆叠面积图与折线图类似,只是图下方的区域全部着色。应用场景有:
- 想要描述数量或体积(而不是价格之类的变量)随时间的变化;
- 有很多数据点。对于很少的数据点,可以考虑绘制柱状图。
- 希望展示各个类别的贡献。
library(ggplot2) library(lubridate) theme_set(theme_bw()) df <- economics[, c("date", "psavert", "uempmed")] df <- df[lubridate::year(df$date) %in% c(1967:1981), ] # labels and breaks for X axis text brks <- df$date[seq(1, length(df$date), 12)] lbls <- lubridate::year(brks) # plot ggplot(df, aes(x=date)) + geom_area(aes(y=psavert+uempmed, fill="psavert")) + geom_area(aes(y=uempmed, fill="uempmed")) + labs(title="Area Chart of Returns Percentage", subtitle="From Wide Data format", caption="Source: Economics", y="Returns %") + # title and caption scale_x_date(labels = lbls, breaks = brks) + # change to monthly ticks and labels scale_fill_manual(name="", values = c("psavert"="#00ba38", "uempmed"="#f8766d")) + # line color theme(panel.grid.minor = element_blank()) # turn off minor grid
堆叠面积图
6.5 日历热力图
当您想要在实际的日历上看到像股票价格这类指标的变化,特别是高点和低点时,日历热力图是一个很好的工具。它强调随着时间的推移视觉上的变化,而不是实际数值的变化。这可以通过使用 geom_tile()
来实现。
拓展:庄小编以前介绍过如何绘制日历图,可参见:calendR包—私人定制专属日历;私人定制日历代码改进。
library(ggplot2) library(plyr) library(scales) library(zoo) df <- read.csv("https://raw.githubusercontent.com/selva86/datasets/master/yahoo.csv") df$date <- as.Date(df$date) # 格式化日期 df <- df[df$year >= 2012, ] # filter reqd years # 创建月周 df$yearmonth <- as.yearmon(df$date) df$yearmonthf <- factor(df$yearmonth) df <- ddply(df,.(yearmonthf), transform, monthweek=1+week-min(week)) # compute week number of month df <- df[, c("year", "yearmonthf", "monthf", "week", "monthweek", "weekdayf", "VIX.Close")] head(df)
ggplot(df, aes(monthweek, weekdayf, fill = VIX.Close)) + geom_tile(colour = "white") + facet_grid(year~monthf) + scale_fill_gradient(low="red", high="green") + labs(x="Week of Month", y="", title = "Time-Series Calendar Heatmap", subtitle="Yahoo Closing Price", fill="Close")
日历热力图
6.6 坡度图
坡度图可以可视化数值和类别排名之间的变化。这更适用于时间点很少的时间序列。下面给出使用 ggplot2 包绘制的案例,来源于:Top 50 ggplot2 Visualizations[5]。内部代码,这里不做过多解释,有能力的读者请自行研究!
此外,关于坡度图的绘制,也有些大佬已经集成 R 包了,例如:CGPfunctions[6] 包中的
newggslopegraph()
、slopegraph [7] 包中的ggslopegraph()
绘制等。
library(dplyr) theme_set(theme_classic()) source_df <- read.csv("https://raw.githubusercontent.com/jkeirstead/r-slopegraph/master/cancer_survival_rates.csv") # 定义函数,来源: https://github.com/jkeirstead/r-slopegraph tufte_sort <- function(df, x="year", y="value", group="group", method="tufte", min.space=0.05) { ## First rename the columns for consistency ids <- match(c(x, y, group), names(df)) df <- df[,ids] names(df) <- c("x", "y", "group") ## Expand grid to ensure every combination has a defined value tmp <- expand.grid(x=unique(df$x), group=unique(df$group)) tmp <- merge(df, tmp, all.y=TRUE) df <- mutate(tmp, y=ifelse(is.na(y), 0, y)) ## Cast into a matrix shape and arrange by first column require(reshape2) tmp <- dcast(df, group ~ x, value.var="y") ord <- order(tmp[,2]) tmp <- tmp[ord,] min.space <- min.space*diff(range(tmp[,-1])) yshift <- numeric(nrow(tmp)) ## Start at "bottom" row ## Repeat for rest of the rows until you hit the top for (i in 2:nrow(tmp)) { ## Shift subsequent row up by equal space so gap between ## two entries is >= minimum mat <- as.matrix(tmp[(i-1):i, -1]) d.min <- min(diff(mat)) yshift[i] <- ifelse(d.min < min.space, min.space - d.min, 0) } tmp <- cbind(tmp, yshift=cumsum(yshift)) scale <- 1 tmp <- melt(tmp, id=c("group", "yshift"), variable.name="x", value.name="y") ## Store these gaps in a separate variable so that they can be scaled ypos = a*yshift + y tmp <- transform(tmp, ypos=y + scale*yshift) return(tmp) } plot_slopegraph <- function(df) { ylabs <- subset(df, x==head(x,1))$group yvals <- subset(df, x==head(x,1))$ypos fontSize <- 3 gg <- ggplot(df,aes(x=x,y=ypos)) + geom_line(aes(group=group),colour="grey80") + geom_point(colour="white",size=8) + geom_text(aes(label=y), size=fontSize, family="American Typewriter") + scale_y_continuous(name="", breaks=yvals, labels=ylabs) return(gg) } ## 准备数据 df <- tufte_sort(source_df, x="year", y="value", group="group", method="tufte", min.space=0.05) df <- transform(df, x=factor(x, levels=c(5,10,15,20), labels=c("5 years","10 years","15 years","20 years")), y=round(y)) ## 绘图 plot_slopegraph(df) + labs(title="Estimates of % survival rates") + theme(axis.title=element_blank(), axis.ticks = element_blank(), plot.title = element_text(hjust=0.5, family = "American Typewriter", face="bold"), axis.text = element_text(family = "American Typewriter", face="bold"))
坡度图
6.7 季节图
如果您正在处理 ts
或 xts
类型的时间序列对象,您可以通过使用 forecast::ggseasonplot()
绘制的季节图来查看季节波动。下面是一个使用 AirPassengers
和 nottem
数据集绘制的例子。
library(ggplot2) library(forecast) theme_set(theme_classic()) # 使用子集数据 nottem_small <- window(nottem, start=c(1920, 1), end=c(1925, 12)) # 使用较小时间窗的子集
nottem 数据集
ggseasonplot(nottem_small) + labs(title="Seasonal plot: Air temperatures at Nottingham Castle")
nottem 数据集的季节图
AirPassengers 数据集
# 绘图 ggseasonplot(AirPassengers) + labs(title="Seasonal plot: International Airline Passengers")
AirPassengers 数据集的季节图
参考资料
[1]
Top 50 ggplot2 Visualizations: http://r-statistics.co/Top50-Ggplot2-Visualizations-MasterList-R-Code.html
[2]
庄闪闪的可视化笔记——常用图形: https://liangliangzhuang.github.io/R-tutorial/main-diagram-types.html
[3]
R Graph Gallery: https://www.r-graph-gallery.com/ggplot2-package.html
[4]
R 语言教程——ggplot 的各种图形: https://www.math.pku.edu.cn/teachers/lidf/docs/Rbook/html/_Rbook/ggplotvis.html
[5]
Top 50 ggplot2 Visualizations: http://r-statistics.co/Top50-Ggplot2-Visualizations-MasterList-R-Code.html
[6]
CGPfunctions: https://cran.r-project.org/web/packages/CGPfunctions/vignettes/Using-newggslopegraph.html
[7]
slopegraph : https://rdrr.io/github/leeper/slopegraph/