Question

我有一些类似于以下内容的数据：

# A tibble: 2,717 x 6
# Groups:   date [60]
   symbol date       monthly.returns score totals score_rank
   <chr>  <date>               <dbl> <dbl>  <dbl>      <int>
 1 GIS    2010-01-29        0.0128   0.436  119.           2
 2 GIS    2010-02-26        0.00982  0.205  120.           1
 3 GIS    2010-03-31       -0.0169   0.549   51.1          3
 4 GIS    2010-04-30        0.0123   0.860   28.0          4
 5 GIS    2010-05-28        0.000984 0.888   91.6          4
 6 GIS    2010-06-30       -0.00267  0.828   15.5          4
 7 GIS    2010-07-30       -0.0297   0.482   81.7          2
 8 GIS    2010-08-31        0.0573   0.408   57.2          3
 9 GIS    2010-09-30        0.0105   0.887   93.3          4
10 GIS    2010-10-29        0.0357   0.111   96.6          1
# ... with 2,707 more rows

我有一个score_rank，我想做的是只要totals列> 100，就可以通过以下方式过滤数据：

1）当score_rank = 1时，根据score列获取观察值的前5％

2）当score_rank = 2或3时，随机抽取5％的观测值

3）当score_rank = 4时，请根据score列减去观察值的5％。

数据：

tickers <- c("GIS", "KR", "MKC", "SJM", "EL", "HRL", "HSY", "K", 
             "KMB", "MDLZ", "MNST", "PEP", "PG", "PM", "SYY", "TAP", "TSN", "WBA", "WMT",
             "MMM", "ABMD", "ACN", "AMD", "AES", "AON", "ANTM", "APA", "CSCO", "CMS", "KO", "GRMN", "GPS",
             "JEC", "SJM", "JPM", "JNPR", "KSU", "KEYS", "KIM", "NBL", "NEM", "NWL", "NFLX", "NEE", "NOC", "TMO", "TXN", "TWTR")

library(tidyquant)
data <- tq_get(tickers,
               get = "stock.prices",              # Collect the stock price data from 2010 - 2015
               from = "2010-01-01",
               to = "2015-01-01") %>%
  group_by(symbol) %>%
  tq_transmute(select = adjusted,                 # Convert the data from daily prices to monthly prices
               mutate_fun = periodReturn,
               period = "monthly",
               type = "arithmetic")

data$score <- runif(nrow(data), min = 0, max = 1)
data$totals <- runif(nrow(data), min = 10, max = 150)

data <- data %>%
  group_by(date) %>%
  mutate(
    score_rank = ntile(score, 4)
  )

编辑：添加了代码。

Answer 1

这是filter的一个选项。为每个相应的“ score_rank”创建一个list个函数（fs），使用map2遍历list个函数和相应的“ score_rank” list vector个，filter（总计）大于100的“数据”和%in%向量的输入的“ score_rank” map2中，应用此函数在“得分”列上filter行样本，并将子集数据与数据filter绑定在一起，其中“总计”小于或等于100

library(purrr)
library(dplyr)
fs <- list(as_mapper(~  . >= quantile(., prob = 0.95)), 
       as_mapper(~ row_number() %in% sample(row_number(), round(0.05 * n()) )),
       as_mapper(~  . <= quantile(., prob = 0.05))
       )


map2_df(list(1, c(2, 3), 4), fs, ~          

    data %>%        
        filter(totals > 100, score_rank %in% .x) %>%
        filter(.y(score))

         )%>%   bind_rows(data %>%
                            filter(totals <= 100))

根据等级和条件过滤数据

1 个答案: