gam
的空间插值Statement
我希望使用通用加性模型(GAM)获得许多空间插值输出。预测单个污染图没有问题,但是,我需要100多个图。如果可能的话,我想实现自动化,并在不超出内存限制的情况下获得结果。
使用GAM(mgcv
程序包]的空间插值过程
仅告诉您,这是获取插值地图的基本步骤。
gam(pollution ~ s(X,Y, k=20))
min
和max
X,Y坐标作为空间范围创建一个空数据框predict
和gam
结果预测空间范围我将展示一个实践方法的实际例子。
举个例子,我创建了一个数据集,如下所示。从df
,您将意识到我有X
Y
和3个污染变量。
library(data.table)
library(mgcv)
X <- c(197745.8,200443.8,200427.6,208213.4,203691.1,208303.0,202546.4,202407.9,202564.8,194095.5,194508.0,195183.8,185432.5,
190249.0,190927.0,197490.1,193551.5,204204.4,199508.4,210201.4,212088.3,191886.5,201045.2,187321.7,205987.0)
Y <- c(451633.1,452496.8,448949.5,449753.3,449282.2,453928.5,452923.2,456347.9,461614.8,456729.3,453019.7,450039.7,449472.0,
444348.1,447274.4,442390.0,443101.2,446446.5,445008.5,446765.2,449508.5,439225.3,460915.6,447392.0,461985.3)
poll1 <- c(34,29,29,33,33,38,35,30,41,43,35,34,41,41,40,36,35,27,53,40,37,32,28,36,33)
poll2 <- c(27,27,34,30,38,36,36,35,37,39,35,33,41,42,40,34,38,31,43,46,38,32,29,33,34)
poll3 <- c(26,30,27,30,37,41,36,36,35,35,35,33,41,36,38,35,34,24,40,43,36,33,30,32,36)
df <- data.table(X, Y, poll1, poll2, poll3)
1。硬编码
如果您看下面的代码,您会意识到我将同一作业复制并粘贴到所有变量中。这将很难实现很多变量。
# Run gam
gam1 <- gam(poll1 ~ s(X,Y, k=20), data = df)
gam2 <- gam(poll2 ~ s(X,Y, k=20), data = df)
gam3 <- gam(poll3 ~ s(X,Y, k=20), data = df)
# "there are over 5000 variables that needs looping
# Create an empty surface for prediction
GAM_poll <- data.frame(expand.grid(X = seq(min(df$X), max(df$X), length=200),
Y = seq(min(df$Y), max(df$Y), length=200)))
# Predict gam results to the empty surface
GAM_poll$gam1 <- predict(gam1, GAM_poll, type = "response")
GAM_poll$gam2 <- predict(gam2, GAM_poll, type = "response")
GAM_poll$gam3 <- predict(gam3, GAM_poll, type = "response")
2。使用for
循环
相反,我列出了清单,并尝试循环所有变量以获取结果。当然,它本身没有问题,但是对多个变量进行迭代将占用所有内存(这是我的经验)。# Run gam using list and for loop
myList <- list()
for(i in 3:length(df)){
myList[[i-2]] <- gam(df[[i]] ~ s(X,Y, k=20), data = df)
}
# Create an empty surface for prediction
GAM_poll <- data.frame(expand.grid(X = seq(min(df$X), max(df$X), length=200),
Y = seq(min(df$Y), max(df$Y), length=200)))
# Predict gam results to the empty surface
myResult <- list()
for(j in 1:length(myList)){
myResult[[j]] <- predict(myList[[j]], GAM_poll, type = "response")
}
寻求帮助
gam
结果?data.table
,purrr
位用户吗?使用gam语句进行空间插值我希望使用广义加性模型(GAM)获得许多空间插值输出。预测单个污染图没有问题,...