在线时间:8:00-16:00
迪恩网络APP
随时随地掌握行业动态
扫描二维码
关注迪恩网络微信公众号
##数据获取 x1=round(runif(100,min=80,max=100)) x2=round(rnorm(100,mean=80, sd=7)) x3=round(rnorm(100,mean=80,sd=18)) x3[which(x3>100)]=100 num=seq(2005138101,length=100) x=data.frame(num,x1,x2,x3) write.table(x, "grade.txt")
##数据分析 y=read.table("grade.txt") mean(y) colMeans(y) colMeans(y)(c("x1","x2","x3")) #//2代表按列取平均值 apply(x, 2, mean) apply(x, 2, max) apply(x, 2, min) apply(x[c("x1","x2","x3")], 2, sum)
#常用分布函数 均匀分布:runif() 正态分布:rnorm() 二项式分布:rbinom() gamma分布:rgamma() 泊松分布:rpois() 指数分布:rexp() 几何分布:rgeom() > which.max(apply(x[c("x1","x2","x3")], 1, sum)) 49 > x$num[which.max(apply(x[c("x1","x2","x3")], 1, sum))] [1] 2005138149 > hist(x$x1) > plot(x$x1,x$x2) > table(x$x1) 80 81 82 83 84 85 86 87 88 89 90 91 92 93 95 96 97 98 99 100 2 3 4 3 8 3 6 6 6 7 4 6 4 8 8 3 5 6 6 2 > barplot(table(x$x1)) > pie(table(x$x2)) > boxplot(x$x1, x$x2, x$x3) > boxplot(x[2:4], col=c("red", "blue", "green"), notch=T) > stars(x[c("x1","x2","x3")]) > stars(x[c("x1","x2","x3")],full=T, draw.segment=T) > faces(x[c("x1","x2","x3")]) > stem(x$x1) The decimal point is at the | 80 | 00000 82 | 0000000 84 | 00000000000 86 | 000000000000 88 | 0000000000000 90 | 0000000000 92 | 000000000000 94 | 00000000 96 | 00000000 98 | 000000000000 100 | 00 > stem(x$x2, scale = 1, width = 80, atom = 1e-08) The decimal point is 1 digit(s) to the right of the | 6 | 23 6 | 7788889 7 | 00001222333444444 7 | 55666677777788888999 8 | 0000000000111122233333344444 8 | 5566666677778999 9 | 001112234 9 | 5 > stem(x$x3, scale = 1, width = 80, atom = 1e-08) The decimal point is 1 digit(s) to the right of the | 3 | 9 4 | 5 | 0022346788 6 | 1112233344555667789 7 | 022223344444555667889 8 | 0122333444455677999 9 | 000112223345666777889 10 | 000000000 > qqnorm(x1) > qqline(x1) > qqnorm(x2) > qqline(x2)
##数据获取 x1=round(runif(100,min=80,max=100)) x2=round(rnorm(100,mean=80, sd=7)) x3=round(rnorm(100,mean=80,sd=18)) x3[which(x3>100)]=100 num=seq(2005138101,length=100) x=data.frame(num,x1,x2,x3) write.table(x, "grade.txt")
##数据分析 y=read.table("grade.txt") mean(y) colMeans(y) colMeans(y)(c("x1","x2","x3")) #//2代表按列取平均值 apply(x, 2, mean) apply(x, 2, max) apply(x, 2, min) apply(x[c("x1","x2","x3")], 2, sum)
#常用分布函数 均匀分布:runif() 正态分布:rnorm() 二项式分布:rbinom() gamma分布:rgamma() 泊松分布:rpois() 指数分布:rexp() 几何分布:rgeom() |
请发表评论