我研究了this question,并基于数千对经度和纬度点创建了自己的4等值线图,但是使用4个等值线中的每个等值线时,我得不到正确的点数上述问题中提到的points.in.polygon方法。
这是到目前为止使用MASS库的代码:
# use kde2d function to create kernel density estimates
x <- pedestrian.df$longitude
y <- pedestrian.df$latitude
dens <- kde2d(x, y, n=200)
# create the contours to plot - 70%, 50%, 25%, 10% of density contained in each contour
prob <- c(0.7, 0.5, 0.25, 0.1)
dx <- diff(dens$x[1:4])
dy <- diff(dens$y[1:4])
sz <- sort(dens$z)
c1 <- cumsum(sz) * dx * dy
levels <- sapply(prob, function(x) {
approx(c1, sz, xout = 1 - x)$y
})
#create the contour plot using smoothScatter which smooths the collisions into kernel densities
smoothScatter(x,y) + contour(dens, levels=levels, labels=prob, col = c("green", "yellow", "orange", "red"), lwd = 1.5, add=T)
这正确地生成了我所期望的:
然后,我尝试使用sp库中的points.in.polygon函数,就像上述链接问题的答案一样:
ls <- contourLines(dens, level=levels)
zone_1 <- point.in.polygon(df$longitude, df$latitude, ls[[4]]$x, ls[[4]]$y)
zone_2 <- point.in.polygon(df$longitude, df$latitude, ls[[3]]$x, ls[[3]]$y)
zone_3 <- point.in.polygon(df$longitude, df$latitude, ls[[2]]$x, ls[[2]]$y)
zone_4 <- point.in.polygon(df$longitude, df$latitude, ls[[1]]$x, ls[[1]]$y)
但这会导致每个区域或轮廓的点数不正确。我知道这是不对的,因为随着轮廓变大,每个轮廓应逐渐具有更多的点。
我尝试查看ls(一个存储多边形的所有x和y坐标的列表的列表),但是有15个级别,而不是我直觉会想到的4个级别。 15个中甚至有多个具有相同值的级别。我怀疑问题的答案在于正确地对列表列表进行细分,以正确地包含与我的4个轮廓相对应的4个级别,但是ls [[1:7]] $ x,ls [[1:7]] $ y没有不行
感谢您的帮助,让我知道是否可以澄清任何事情!
答案 0 :(得分:2)
我认为 pedestrian
是您自己的数据,而不是pkg中的数据,由于这不是问题的一部分,因此我们将使用另一种数据:
library(MASS)
library(sp)
attach(geyser)
data.frame(
x = geyser$duration,
y = geyser$waiting
) -> xdf
dens <- kde2d(xdf$x, xdf$y, n = 100)
prob <- c(0.7, 0.5, 0.25, 0.1)
dx <- diff(dens$x[1:4])
dy <- diff(dens$y[1:4])
sz <- sort(dens$z)
c1 <- cumsum(sz) * dx * dy
levels <- sapply(prob, function(x) {
approx(c1, sz, xout = 1 - x)$y
})
smoothScatter(x,y) +
contour(dens, levels=levels, labels=prob, col = c("green", "yellow", "orange", "red"), lwd = 1.5, add=TRUE)
“多个级别”的原因是,给定图层中的每个多边形都是独立的,因此每个级别可能> 1个:
cl <- contourLines(dens, level=levels)
sort(table(sapply(cl, `[[`, "level")))
## 0.00519851181336958 0.00765971436995347 0.0107843979424224 0.0128423136194731
## 2 3 3 3
因此,在计算每个多边形的点数时只需考虑这一点:
setNames(
lapply(cl, function(poly) sum(sp::point.in.polygon(xdf$x, xdf$y, poly$x, poly$y))),
sapply(cl, `[[`, "level")
) -> level_cts
str(level_cts)
## List of 11
## $ 0.00519851181336958: int 91
## $ 0.00519851181336958: int 174
## $ 0.00765971436995347: int 78
## $ 0.00765971436995347: int 57
## $ 0.00765971436995347: int 74
## $ 0.0107843979424224 : int 65
## $ 0.0107843979424224 : int 34
## $ 0.0107843979424224 : int 33
## $ 0.0128423136194731 : int 42
## $ 0.0128423136194731 : int 10
## $ 0.0128423136194731 : int 3
然后我们可以总结一下:
sapply(
split(level_cts, names(level_cts)),
function(level) sum(unlist(level))
) -> pt_cts
pt_cts
## 0.00519851181336958 0.00765971436995347
## 265 209
## 0.0107843979424224 0.0128423136194731
## 132 55
然后,获取%:
pt_cts / nrow(xdf)
## 0.00519851181336958 0.00765971436995347
## 0.8862876 0.6989967
## 0.0107843979424224 0.0128423136194731
## 0.4414716 0.1839465
更新
我们不仅可以计算百分比,还可以将级别分配给原始数据:
do.call(
rbind.data.frame,
lapply(cl, function(poly) { # iterate over each polygon
# figure out which pts are in this polgyon
which_pts <- as.logical(sp::point.in.polygon(xdf$x, xdf$y, poly$x, poly$y))
tdf <- xdf[which_pts,] # assign them to a temp data frame
tdf$level <- poly$level # add the level
tdf
})
) -> new_xdf
dplyr::glimpse(new_xdf)
## Observations: 661
## Variables: 3
## $ x <dbl> 2.000000, 2.033333, 1.833333, 1.616667, 1.766667, 2.0000...
## $ y <dbl> 77, 77, 81, 89, 73, 83, 84, 85, 79, 75, 91, 87, 86, 78, ...
## $ level <dbl> 0.005198512, 0.005198512, 0.005198512, 0.005198512, 0.00...
# while you likely want the level value, this adds columns for level # & prob
new_xdf$level_num <- as.integer(factor(new_xdf$level, levels, labels=1:length(levels)))
new_xdf$prob <- as.numeric(as.character(factor(new_xdf$level, levels, labels=prob)))
dplyr::glimpse(new_xdf)
## Observations: 661
## Variables: 5
## $ x <dbl> 2.000000, 2.033333, 1.833333, 1.616667, 1.766667, 2....
## $ y <dbl> 77, 77, 81, 89, 73, 83, 84, 85, 79, 75, 91, 87, 86, ...
## $ level <dbl> 0.005198512, 0.005198512, 0.005198512, 0.005198512, ...
## $ level_num <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...
## $ prob <dbl> 0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0....
dplyr::count(new_xdf, level, level_num, prob)
## # A tibble: 4 x 4
## level level_num prob n
## <dbl> <int> <dbl> <int>
## 1 0.00520 1 0.700 265
## 2 0.00766 2 0.500 209
## 3 0.0108 3 0.250 132
## 4 0.0128 4 0.100 55