Insert rows with zeros in data frames in R [duplic

问题:

This question already has an answer here:

• Filling missing dates by group 3 answers

Consider a fragmented dataset like this:

   ID       Date Value
1   1 2012-01-01  5065
4   1 2012-01-04  1508
5   1 2012-01-05  9489
6   1 2012-01-06  7613
7   2 2012-01-07  6896
8   2 2012-01-08  2643
11  3 2012-01-02  7294
12  3 2012-01-03  8726
13  3 2012-01-04  6262
14  3 2012-01-05  2999
15  3 2012-01-06 10000
16  3 2012-01-07  1405
18  3 2012-01-09  8372

Notice that observations are missing for (2,3,9,10,17). What I would like, is to fill out some of these gaps in the dataset with "Value" = 0, like so:

   ID       Date Value
1   1 2012-01-01  5920
2   1 2012-01-02     0
3   1 2012-01-03     0
4   1 2012-01-04  8377
5   1 2012-01-05  7810
6   1 2012-01-06  6452
7   2 2012-01-07  3483
8   2 2012-01-08  5426
9   2 2012-01-09     0
11  3 2012-01-02  7854
12  3 2012-01-03  1948
13  3 2012-01-04  7141
14  3 2012-01-05  5402
15  3 2012-01-06  6412
16  3 2012-01-07  7043
17  3 2012-01-08     0
18  3 2012-01-09  3270

The point is that the zeros only should be inserted if there is a past observation for the same (grouped) ID. I would like to avoid any loops, as the full dataset is quite large.

Any suggestions? To reproduce the dataframe:

df <- data.frame(matrix(0, nrow = 18, ncol = 3,
                  dimnames = list(NULL, c("ID","Date","Value"))) )
df[,1] = c(1,1,1,1,1,1,2,2,2,3,3,3,3,3,3,3,3,3) 
df[,2] = seq(as.Date("2012-01-01"),
             as.Date("2012-01-9"), 
             by=1)
df[,3] = sample(1000:10000,18,replace=T)
df = df[-c(2,3,9,10,17),]

回答1:

There are already some solid answers here, but I would recommend checking out the package padr.

library(dplyr)
library(padr)

df %>% 
  pad(start_val = as.Date("2012-01-01"),
      end_val =   as.Date("2012-01-09"),
      group = "ID") %>% 
  fill_by_value(Value)

The package gives some pretty intuitive functions for summarizing Date columns as well.

回答2:

Tidyverse has complete which is a nice easy way to expand something like this. We can also use the fill argument to replace the NAs with zero in the same step.

library(tidyverse)

df %>% group_by(ID) %>% 
  complete(Date = seq(min(Date), max(Date), "day"), fill = list(Value = 0)) 

# A tibble: 16 x 3
# Groups:   ID [3]
      ID Date       Value
   <dbl> <date>     <dbl>
 1     1 2012-01-01  1047
 2     1 2012-01-02     0
 3     1 2012-01-03     0
 4     1 2012-01-04  8147
 5     1 2012-01-05  1359
 6     1 2012-01-06  1892
 7     2 2012-01-07  3362
 8     2 2012-01-08  8988
 9     3 2012-01-02  2731
10     3 2012-01-03  9794

...

回答3:

The following is a base R solution. It uses split to divide the input into sub-dataframes and then lapply to process each of them.

result <- lapply(split(df, df$ID), function(DF){
  Date <- seq(min(DF$Date), max(DF$Date), by = "days")
  DF2 <- data.frame(ID = rep(DF$ID[1], length.out = length(Date)))
  DF2$Date <- Date
  DF2$Value <- 0
  DF2$Value[Date %in% DF$Date] <- DF$Value
  DF2
})

result <- do.call(rbind, result)
row.names(result) <- NULL
result