如何在R中找到线性趋势？

Question

我有一个大约有160个数据点的矩阵。我有兴趣发现该系列的任何部分（不必是整个系列）都有趋势（线性和增加/减少）。我只对找出线性趋势而非非线性趋势感兴趣。

我首先在表格的基本数据集上进行测试：

    A<-c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,
16,17,17,18,17,17,18,17,17,18,19,19,19,19,19,19,19,19,19,19,19)

这里只是绘制数据，我认为趋势是从1:17向上，我想为此引入假人（一旦我知道趋势起点和终点，我就能做到）。有什么方法可以知道吗？ IN R？

Answer 1

A<-c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,
     16,17,17,18,17,17,18,17,17,18,19,19,19,19,19,19,19,19,19,19,19)
x <- seq_along(A)

plot(A ~ x)

fit0 <- lm(A ~ x)

library(segmented)
fit1 <- segmented(fit0, seg.Z = ~ x, psi = list(x = c(10, 40)))
summary(fit1)
#Estimated Break-Point(s):
#         Est. St.Err
#psi1.x 18.00  0.220
#psi2.x 34.04  0.246


lines(x, predict(fit1), col = "red")

Answer 2

检查lm功能。此功能可帮助您创建线性回归。

A<-c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,17,18,17,17,18,17,17,18,19,19,19,19,19,19,19,19,19,19,19)
x = c(1:length(A))  
plot(x,A)
reg = lm(A~x)
abline(reg,col="red")

如果您想在 startPoint 开始线性回归，请使用“A [ startPoint ：length（A）]

了解趋势：print(reg)

Answer 3

分解功能可以提供帮助。假人可能不够灵敏。您可能只想减去趋势分量。尝试：

A<-c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,
 16,17,17,18,17,17,18,17,17,18,19,19,19,19,19,19,19,19,19,19,19)

# convert to ts first
my.ts <- ts(A, start = c(2000, 1), freq = 12) # .. if you have months
my.trend <- decompose(my.ts)$trend
print(max(diff(my.trend), na.rm = TRUE))

# use a certain percentage of max difference to find start and end of trend
tolerance <- 0.1 # say 10 % ..
trend.start <- which.max(my.trend >= tolerance * max(my.trend, na.rm=T))
trend.end <- which.max(my.trend >= (1-tolerance) * max(my.trend, na.rm=T))

# plot shows that it - almost - fits
plot(my.ts, col = "blue", type = "o")
abline(v = time(my.ts)[trend.start], col = "gray") 
abline(v = time(my.ts)[trend.end], col = "gray")
lines(my.trend)

# but what do you do, if you have a ts like this ..
B <- c(rep(0,50), seq(1,30,0.5), rep(30,50), seq(30,1,-1)) + rnorm (189,0,1)
my.ts <- ts(B, start = c(2000, 01), freq = 12)
plot(my.ts)
my.trend <- decompose(my.ts)$trend
plot(my.trend)