Question

在使用MICE软件包对某些数据集进行多次插补之后，我想为两个因变量（得分1，得分_2）中的每一个分别计算线性回归模型。两个模型的自变量（手臂，性别，年龄，基线得分）相同。不幸的是，我无法管理（1）将mouse :: pool（）函数集成到dplyr管道中，以及（2）我不知道如何通过因变量对mouse :: pool（）函数进行分组（ df_lm $ score）。

library(tidyverse)
library(mice)


# SIMULATE DATA


df <- data.frame(id = 1:120,
                 arm = sample(c('intervention', 'control'), 120, replace = TRUE),
                 sex = sample(c('m', 'f'), 120, replace = TRUE),
                 age = round(rnorm(120, 55, 10)),
                 score_1 = round(rnorm(120, 50, 5)),
                 score_2 = round(rnorm(120, 50, 7)))

df <- df %>% bind_rows(df) %>%
  mutate(time = c(rep('baseline', 120), rep('follow_up', 120))) %>%
  select(id, arm, time, everything()) %>%
  gather(score, measure , -(id:age)) %>%
    spread(key = time, value = measure)

# INSERT SOME MISSING VALUES

df$follow_up[seq(1, 240, 5)] <- NA

# IMPUTATION MODEL 

init <- mice(df, maxit = 0)
predM <- init$predictorMatrix
# remove as predictor
predM[ , c('arm')] <- 0


mids_from_df <- mice(df,
  method = 'pmm',
  predictorMatrix = predM,
  m = 5,
  seed = 123,
  print = FALSE
)


# COMPUTE MODELS


fmla <- "follow_up ~ baseline + arm + sex + age"

df_lm <- mids_from_df %>%
  mice::complete("long", include = FALSE) %>%
  group_by(.imp, score) %>%
  nest() %>%
  mutate(lm_model = map(data, ~lm(fmla, data = .)))

我想分别获取每个因变量的合并结果。但是，我不知道如何与dplyr和purr一起使用mouse :: pool（）。以下代码引发错误：

df_lm <- mids_from_df %>%
  mice::complete("long", include = FALSE) %>%
  group_by(.imp, score) %>%
  nest() %>%
  mutate(lm_model = map(data, ~lm(fmla, data = .)))  %>%
  group_by(.imp, score) %>%
  pool(., lm_model) # does not work

错误消息是：“错误：类整数对象的无视方法”

在此先感谢您的帮助！

可以使用dplyr对MICE pool（）的输出进行分组吗？

0 个答案: