data scientist, analyst <- user of c(R, Python, spark(scala, R, Python), hadoop)
getwd()
light <- read.table("C:\\Users\\park\\therbook\\light.txt", header = T)
light
hist(light$speed, col = "green")
summary(light)
boxplot(light)
x <- exp(rnorm(1000))
hist(x)
qqnorm(light$speed)
qqline(light$speed, col = "red") # 빨간 선 안에 올라가야 정규분포를 따른다고 봄.
H0: normally distributed => not reject
shapiro.test(light$speed) # normality test p-value > 0.05 따라서 H0 기각 힘듦
qqnorm(x)
qqline(x, col = "red")
H0: normally distributed => reject
shapiro.test(x)
H0 : speed of light 299 990 km/s => reject
wilcox.test(light$speed,mu=990)
H0 : speed of light 299 909 km/s => not reject
wilcox.test(light$speed,mu=909)