Geek asks and answers

Delete cases, all identifiers, for cases when changes in NO occurred between time 1 and time 2

I have two data frames with many cases. One of time 1 and one of time 2. I am looking for a way to quickly identify cases where changes have occurred between time 1 and time 2, and I'm a bit stuck.

Here is an example. So, I have a data frame from time 1,

df.t1 <- data.frame(id = c(1,1,1,2,2,3,3,5,5,6), ABC = LETTERS[1:10], Num = 101:110)

and looks like this:

df.t1
   id ABC Num
1   1   A 101
2   1   B 102
3   1   C 103
4   2   D 104
5   2   E 105
6   3   F 106
7   3   G 107
8   5   H 108
9   5   I 109
10  6   J 110

time when two rolls around

df.t2 <- df.t1

and there are some changes

df.t2[3,3] <- 104
df.t2[2,2] <- "H"
df.t2[8,3] <- 999
df.t2[10,3] <- NA
df.t2[11,] <- c(3, "J", 107)

it's time 2,

df.t2
   id ABC  Num
1   1   A  101
2   1   H  102
3   1   C  104
4   2   D  104
5   2   E  105
6   3   F  106
7   3   G  107
8   5   H  999
9   5   I  109
10  6   J <NA>
11  3   J  107

Now I’m looking for a quick way to remove cases, all identifiers, for cases when in case (NO) there were NO changes (any line) between time1 and time 2. In a specific example, it is only with id # 2 that no changes have occurred between time 1 and time 2.

I am looking for an end result that looks like this:

(df <- subset(df.t2, id != 2))
   id ABC  Num
1   1   A  101
2   1   H  102
3   1   C  104
6   3   F  106
7   3   G  107
8   5   H  999
9   5   I  109
10  6   J <NA>
11  3   J  107

Any help would be appreciated.

+5
5

@alexwhan fooobar.com/questions/1145579/..., , - . , .

# get a matrix of unique integers for each column (stacking the two data frames)
ms <- do.call(cbind, lapply(seq_len(ncol(df.t1)), function(ni) {
  xi <- c(as.character(df.t1[[ni]]), as.character(df.t2[[ni]]))
  match(xi, unique(xi))
}))
# convert to base max(ms) to get a single unique identifying number
us <- as.vector(ms %*% max(ms)^c(0:(ncol(ms)-1)))
u2 <- us[(nrow(df.t1)+1):length(us)]
u1 <- us[1:nrow(df.t1)]
# now get changed values and proceed as in alexwhan answer
ch <- unique(df.t2$id[! u2 %in% u1])
df.t2[df.t2$id %in% ch,]

us, , , , , , , . ( ms, , .)

base <- as.integer(max(ms)^c(0:(nrow(ms)-1)))
us <- apply((ms-1L) * base, 2, sum)
+2

, . data.table:

require(data.table)
dt1 <- data.table(df.t1)
dt2 <- data.table(df.t2)
# your conversion made them characters
dt2[, `:=`(id = as.numeric(id), Num = as.numeric(Num))]
setkey(dt1, "id", "ABC")
setkey(dt2, "id", "ABC")
dt <- dt1[dt2]
dt2[id %in% dt[, !(all(Num == Num.1)) | any(is.na(c(Num, Num.1))), by=id][V1 == TRUE]$id]

#    id ABC Num
# 1:  1   A 101
# 2:  1   C 104
# 3:  1   H 102
# 4:  3   F 106
# 5:  3   G 107
# 6:  3   J 107
# 7:  5   H 999
# 8:  5   I 109
# 9:  6   J  NA

, dt = dt1[dt2]:

dt2[id %in% dt[, ctr := Num %in% Num.1, by=1:nrow(dt)][ctr == FALSE, unique(id)]]
+3

How about using data.table and inputting everything id, ABCandNUM

require(data.table)
dt1 <- data.table(df.t1)
dt2 <- data.table(df.t2)
# your conversion made them characters
dt2[, `:=`(id = as.numeric(id), Num = as.numeric(Num))]
setkey(dt1, "id", "ABC", "Num")
setkey(dt2, "id", "ABC", "Num")
# then it is just

dt2[dt2[!dt1][,list(unique(id))]]


   id ABC Num
1:  1   A 101
2:  1   C 104
3:  1   H 102
4:  3   F 106
5:  3   G 107
6:  3   J 107
7:  5   H 999
8:  5   I 109
9:  6   J  NA

This uses a non-unit connection between dt2 and dt1, selects unique id values ​​from them, and then multiplies the data accordingly dt2.

with caution about NA values ​​in keys that are problematic ....

+3
source

Try the following:

df.t1$interact <- with(df.t1, interaction(id, ABC, Num))
df.t2$interact <- with(df.t2, interaction(id, ABC, Num))

change.ids <- unique(df.t2$id[!df.t2$interact %in% df.t1$interact])
new.df <- df.t2[df.t2$id %in% change.ids,]
+2
source

OK, here is an alternative with some loop. I have not tried this with a more complex example, but I'll see:

no.change <- vector()
#identifies rows which don't change
for(i in 1:nrow(df.t2)) { 
    for(k in 1:nrow(df.t1)) {
    if(all(df.t2[i,]==df.t1[k,],na.rm=TRUE)) #na.rm gets rid of error
   no.change <- c(no.change, i)
  }
}
# gets ids from all the other rows
id.change <- df.t2$id[-no.change]
df <- df.t2[df.t2$id %in% id.change,]
+1
source

More articles:


All Articles