使用具有 NA 的数据框中的多个类别按行计算斜率和相关统计数据

Question

我查看了在“How can I calculate the slope of multiple subsets of a data frame more efficiently?”下发布的相关问题，但我的初学者身份不允许我直接对该主题发表评论（不知道该怎么做），所以我在这里问：

如何使用 dplyr&broom 包解决方案避免数据集中的 NA 处理数据框中多个类别的斜率计算？这是脚本和结果的示例？

示例数据：

DOY<-c(102,102,102,102,102,102,102,102,102,102,212,212,212,212,212,212, 212,212,212,212)
LOCATION <- c(1,1,1,1,1,2,2,2,2,2,1,1,1,1,1,3,3,3,3,3)
response <-c(NA,NA,NA,NA,NA,7,10,15,20,30,2,4,6,NA,8,10,15,20,30,NA) 
ts <- c(0,10,20,30, 40,0,10,20,30,40,0,10,20,30,40,0,10,20,30,40)
test.data <- data.frame(cbind(DOY, LOCATION, response, ts))

    library(dplyr)
    library(broom) 

 test.data2 <- test.data %>%  group_by(DOY) %>% do(tidy(lm(response ~ ts, data = .))) 
   test.data2 %>% filter(term == "ts")

一个条件的结果有效（因为每行有足够的数据，没有 NA）：

# A tibble: 2 x 6
   # Groups:   DOY [2]
   #            DOY    term  estimate    std.error   statistic   p.value
   #            <dbl>  <chr>    <dbl>     <dbl>       <dbl>       <dbl>
   #     1     102.     ts       0.560    0.0721      7.77      0.00444
   #     2     212.     ts       0.278    0.247       1.13      0.303

但如果使用多个类别进行分组，则不会：

test.dataX <- test.data %>%  group_by(LOCATION, DOY) %>% do(tidy(lm(response ~ ts, data = .)))

错误结果：

   # Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, ...) : 
   #  0 (non-NA) cases

 test.dataX %>% filter(term == "ts")
   # Error in eval(lhs, parent, parent) : object 'test.dataX' not found

尝试 2：我在 lm() 中尝试了 na.omit，但这也没有用：

test.dataX <- test.data %>%  group_by(LOCATION, DOY) %>% do(tidy(lm(response ~ ts, data = ., na.action=na.omit)))
   # Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, ...) : 
   #  0 (non-NA) cases

理想情况下，我希望有这样的东西（如果可能，连同 R2 - 如何将其添加到上面的输出中）？

# DOY   LOCATION    slope     R2
 # 102    1         NA        NA
 # 102    2         0.560     0.953
 # 212    1         0.149     0.966
 # 212    3         0.650     0.966

########################

请多多指教。谢谢！

Answer 1

如果我们想return NA，那么使用possibly

library(tidyverse)
library(broom)
pos1 <-  possibly(lm, otherwise = NULL)
prsq <- possibly(pull, otherwise = NA)
test.data %>%
     group_by(DOY, LOCATION) %>%
     nest %>%
     mutate(model = map(data, ~ pos1(response~ ts, data = .x)),
            slope = map_dbl(model, ~ 
                            .x %>% 
                                tidy %>%
                                filter(term == 'ts') %>%
                                prsq(estimate)),
            R2 = map_dbl(model, ~ 
                             .x %>%
                                   glance %>%
                                   prsq(r.squared))) %>%
      select(-data, -model)
# A tibble: 4 x 4
#    DOY LOCATION  slope     R2
#  <dbl>    <dbl>  <dbl>  <dbl>
#1   102        1 NA     NA    
#2   102        2  0.56   0.953
#3   212        1  0.149  0.966
#4   212        3  0.650  0.966

使用具有 NA 的数据框中的多个类别按行计算斜率和相关统计数据

Calculating slope and associated stats by rows using multiple categories in dataframe that has NAs

r

dplyr

tidyr

lm

broom