5天仍未回答
我一直在努力解决这个问题,任何帮助都会非常感激。我正在尝试编写一个运行几个逐步回归的函数,并将所有这些函数输出到列表中。但是,R在读取我在函数参数中指定的数据集时遇到问题。我在各个主板(here,here和here)上发现了几个类似的错误,但是它们似乎都没有得到解决。这一切都归结为在用户定义的函数中调用step()的一些奇怪问题。我使用以下脚本来测试我的代码。多次运行整个过程,直到出现错误(相信我,它会):
test.df <- data.frame(a = sample(0:1, 100, rep = T),
b = as.factor(sample(0:5, 100, rep = T)),
c = runif(100, 0, 100),
d = rnorm(100, 50, 50))
test.df$b[10:100] <- test.df$a[10:100] #making sure that at least one of the variables has some predictive power
stepModel <- function(modeling.formula, dataset, outfile = NULL) {
if (is.null(outfile) == FALSE){
sink(file = outfile,
append = TRUE, type = "output")
print("")
print("Models run at:")
print(Sys.time())
}
model.initial <- glm(modeling.formula,
family = binomial,
data = dataset)
model.stepwise1 <- step(model.initial, direction = "backward")
model.stepwise2 <- step(model.stepwise1, scope = ~.^2)
output <- list(modInitial = model.initial, modStep1 = model.stepwise1, modStep2 = model.stepwise2)
sink()
return(output)
}
blah <- stepModel(a~., dataset = test.df)
这会返回以下错误消息(如果错误没有立即显示,请继续重新运行test.df脚本以及调用stepModel(),它最终会显示):
Error in is.data.frame(data) : object 'dataset' not found
我已经确定一切都运行正常,直到model.stepwise2开始构建。不知何故,临时对象'数据集'在第一步逐步回归中工作正常,但第二步无法识别。我通过评论部分功能找到了这一点,如下所示。此代码运行正常,证明对象'dataset'最初被识别:
stepModel1 <- function(modeling.formula, dataset, outfile = NULL) {
if (is.null(outfile) == FALSE){
sink(file = outfile,
append = TRUE, type = "output")
print("")
print("Models run at:")
print(Sys.time())
}
model.initial <- glm(modeling.formula,
family = binomial,
data = dataset)
model.stepwise1 <- step(model.initial, direction = "backward")
# model.stepwise2 <- step(model.stepwise1, scope = ~.^2)
# sink()
# output <- list(modInitial = model.initial, modStep1 = model.stepwise1, modStep2 = model.stepwise2)
return(model.stepwise1)
}
blah1 <- stepModel1(a~., dataset = test.df)
编辑 - 在有人询问之前,所有的summary()函数都在那里,因为完整的函数(我编辑它以便你可以专注于错误)有另一个片段定义一个文件到你可以输出逐步跟踪。我摆脱了他们
编辑2 - 会话信息
sessionInfo() R版本2.15.1(2012-06-22) 平台:x86_64-pc-mingw32 / x64(64位)
locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
[5] LC_TIME=English_United States.1252
attached base packages:
[1] tcltk stats graphics grDevices utils datasets methods base
other attached packages:
[1] sqldf_0.4-6.4 RSQLite.extfuns_0.0.1 RSQLite_0.11.3 chron_2.3-43
[5] gsubfn_0.6-5 proto_0.3-10 DBI_0.2-6 ggplot2_0.9.3.1
[9] caret_5.15-61 reshape2_1.2.2 lattice_0.20-6 foreach_1.4.0
[13] cluster_1.14.2 plyr_1.8
loaded via a namespace (and not attached):
[1] codetools_0.2-8 colorspace_1.2-1 dichromat_2.0-0 digest_0.6.2 grid_2.15.1
[6] gtable_0.1.2 iterators_1.0.6 labeling_0.1 MASS_7.3-18 munsell_0.4
[11] RColorBrewer_1.0-5 scales_0.2.3 stringr_0.6.2 tools_2.15
编辑3 - 执行与该功能相同的操作,只是不使用函数。即使算法没有收敛,每次运行也都会正常运行:
modeling.formula <- a~.
dataset <- test.df
outfile <- NULL
if (is.null(outfile) == FALSE){
sink(file = outfile,
append = TRUE, type = "output")
print("")
print("Models run at:")
print(Sys.time())
}
model.initial <- glm(modeling.formula,
family = binomial,
data = dataset)
model.stepwise1 <- step(model.initial, direction = "backward")
model.stepwise2 <- step(model.stepwise1, scope = ~.^2)
output <- list(modInitial = model.initial, modStep1 = model.stepwise1, modStep2 = model.stepwise2)
答案 0 :(得分:5)
使用do.call
来引用调用环境中的数据集对我有用。有关原始建议,请参阅https://stackoverflow.com/a/7668846/210673。这是一个有效的版本(删除了sink
代码)。
stepModel2 <- function(modeling.formula, dataset) {
model.initial <- do.call("glm", list(modeling.formula,
family = "binomial",
data = as.name(dataset)))
model.stepwise1 <- step(model.initial, direction = "backward")
model.stepwise2 <- step(model.stepwise1, scope = ~.^2)
list(modInitial = model.initial, modStep1 = model.stepwise1, modStep2 = model.stepwise2)
}
blah <- stepModel2(a~., dataset = "test.df")
对于set.seed(6)
与原始代码一致,它失败了。它失败的原因是dataset
函数中不存在step
变量,虽然在model.stepwise1
中不需要它,但model.stepwise2
model.stepwise1
时需要它{{1}} 1}}保持一个线性项。因此,当您的版本失败时就是这种情况。像我一样在全球环境中调用数据集解决了这个问题。