group_by 和计算 durbin watson 测试

Question

假设我有以下数据。我想为我的时间序列的 eahc 计算 Durbin Watson 测试。然后我试图在箱线图中绘制每家公司的 DW 值。

library(tidyquant)
library(lmtest)

data(FANG)

FANG_annual_returns <- FANG %>%
  group_by(symbol) %>%
  tq_transmute(select     = adjusted, 
               mutate_fun = periodReturn, 
               period     = "daily", 
               type       = "arithmetic")

FANG_annual_returns

FANG_annual_returns %>%
  group_by(symbol) %>%
  summarise(res = dwtest(daily.returns))

我有点卡在summarise、do、map 部分。

提前致谢！

Answer 1

这是一个使用tidyr::nest 和map 的选项。我使用daily.returns~1 作为dwtest 和tidy::broom 的公式以获得干净的输出。

library(tidyquant)
library(lmtest)
library(broom)
FANG_annual_returns %>%
     nest(-symbol) %>%
     mutate(res = map(data, ~broom::tidy(dwtest(daily.returns~1, data=.x)))) %>% 
     unnest(res)

# A tibble: 4 x 6
   symbol data                 statistic  p.value method             alternative                           
    <chr>  <list>                   <dbl>    <dbl> <chr>              <chr>                                 
  1 FB     <tibble [1,008 x 2]>      1.98 0.402    Durbin-Watson test true autocorrelation is greater than 0
  2 AMZN   <tibble [1,008 x 2]>      1.98 0.402    Durbin-Watson test true autocorrelation is greater than 0
  3 NFLX   <tibble [1,008 x 2]>      1.80 0.000740 Durbin-Watson test true autocorrelation is greater than 0
  4 GOOG   <tibble [1,008 x 2]>      1.93 0.143    Durbin-Watson test true autocorrelation is greater than 0