R running average for non-time data

≯℡__Kan透↙ 提交于 2019-12-01 00:37:30

One way to provide a running mean is with geom_smooth() using the loess local regression method. In order to demonstrate my proposed solution, I created a fake genomic dataset using R functions. You can adjust the span parameter of geom_smooth to make the running mean smoother (closer to 1.0) or rougher (closer to 1/number of data points).

# Create example data.
set.seed(27182)

y1 = rnorm(10000) + 
     c(rep(0, 1000), dnorm(seq(-2, 5, length.out=8000)) * 3, rep(0, 1000))
y2 = c(rnorm(2000), rnorm(1000, mean=1.5), rnorm(1000, mean=-1, sd=2), 
       rnorm(2000, sd=2))
y3 = rnorm(4000)
pos = c(sort(runif(10000, min=0, max=1e8)),
        sort(runif(6000,  min=0, max=6e7)),
        sort(runif(4000,  min=0, max=4e7)))
chr = rep(c("chr01", "chr02", "chr03"), c(10000, 6000, 4000))

data1 = data.frame(CHROM=chr, POS=pos, DIFF=c(y1, y2, y3))

# Plot.
p = ggplot(data1, aes(x=POS, y=DIFF)) +
    geom_point(alpha=0.1, size=1.5) +
    geom_smooth(colour="darkgoldenrod1", size=1.5, method="loess", degree=0, 
        span=0.1, se=FALSE) +
    scale_x_continuous(breaks=seq(1e7, 3e8, 1e7), 
        labels=paste(seq(10, 300, 10)), expand=c(0, 0)) +
    xlab("Position, Megabases") +
    theme(axis.text.x=element_text(size=8)) +
    facet_grid(. ~ CHROM, scales="free", space="free")

ggsave(filename="plot_1.png", plot=p, width=10, height=5, dpi=150)

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!