Related
I have data as follows:
DT <- structure(list(Area = c("A", "A", "A", "A", "B", "B", "B", "B"
), Year = c(1, 1, 2, 2, 1, 1, 2, 2), Group = c(1, 2, 1, 2, 1,
2, 1, 2), Population_Count = c(10, 12, 10, 12, 10, 13, 10, 11
), Male_Count = c(5, 7, 5, 4, 5, 8, 5, 6), Female_Count = c(5,
5, 5, 8, 5, 5, 5, 5)), row.names = c(NA, -8L), class = c("tbl_df",
"tbl", "data.frame"))
# A tibble: 8 x 6
Area Year Group Population_Count Male_Count Female_Count
<chr> <dbl> <dbl> <dbl> <dbl> <dbl>
1 A 1 1 10 5 5
2 A 1 2 12 7 5
3 A 2 1 10 5 5
4 A 2 2 12 4 8
5 B 1 1 10 5 5
6 B 1 2 13 8 5
7 B 2 1 10 5 5
8 B 2 2 11 6 5
I would like to keep one observations per Area-Year, without losing any information. I tried to do
DTcast <- dcast(DT, Area + Year ~ Group + Population_Count + Male_Count + Female_Count)
But that results in a lot of rubbish:
Area Year 1_10_5_5 2_11_6_5 2_12_4_8 2_12_7_5 2_13_8_5
1 A 1 5 NA NA 5 NA
2 A 2 5 NA 8 NA NA
3 B 1 5 NA NA NA 5
4 B 2 5 5 NA NA NA
In addition, when I apply it to the actual data, I get:
Using 'H_FEMALE' as value column. Use 'value.var' to override
Error in CJ(1:72284, 1:1333365) :
Cross product of elements provided to CJ() would result in 96380955660 rows which exceeds .Machine$integer.max == 2147483647
So I think I am doing something wrong. I think it maybe has to do with the value.var which I do not know how to select.
Desired result:
# A tibble: 4 x 9
Area Year Group `Population_Count_ Group_1` `Male_Count_ Group_1` `Female_Count_ Group_1` `Population_Count_ Group_2` `Male_Count_ Group_2` `Female_Count_ Group_2`
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 A 1 1 10 5 5 12 7 5
2 A 2 1 10 5 5 12 4 8
3 B 1 1 10 5 5 13 8 5
4 B 2 1 10 5 5 11 6 5
library(tidyverse)
DT %>% pivot_wider(id_cols = c("Area", "Year"), names_from = "Group", values_from = 4:6)
> DT %>% pivot_wider(id_cols = c("Area", "Year"), names_from = "Group", values_from = 4:6)
# A tibble: 4 x 8
Area Year Population_Count_1 Population_Count_2 Male_Count_1 Male_Count_2 Female_Count_1 Female_Count_2
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 A 1 10 12 5 7 5 5
2 A 2 10 12 5 4 5 8
3 B 1 10 13 5 8 5 5
4 B 2 10 11 5 6 5 5
This will name your columns as desired
DT %>% pivot_wider(id_cols = c("Area", "Year"),
names_from = "Group",
values_from = 4:6,
names_sep = "_Group_")
use data.table
library(data.table)
dt <- structure(list(Area = c("A", "A", "A", "A", "B", "B", "B", "B"
), Year = c(1, 1, 2, 2, 1, 1, 2, 2), Group = c(1, 2, 1, 2, 1,
2, 1, 2), Population_Count = c(10, 12, 10, 12, 10, 13, 10, 11
), Male_Count = c(5, 7, 5, 4, 5, 8, 5, 6), Female_Count = c(5,
5, 5, 8, 5, 5, 5, 5)), row.names = c(NA, -8L), class = c("tbl_df",
"tbl", "data.frame"))
setDT(dt)
dcast(
dt,
formula = Area + Year ~ Group,
value.var = grep("_Count", names(dt), value = T)
)
#> Area Year Population_Count_1 Population_Count_2 Male_Count_1 Male_Count_2
#> 1: A 1 10 12 5 7
#> 2: A 2 10 12 5 4
#> 3: B 1 10 13 5 8
#> 4: B 2 10 11 5 6
#> Female_Count_1 Female_Count_2
#> 1: 5 5
#> 2: 5 8
#> 3: 5 5
#> 4: 5 5
Created on 2020-12-18 by the reprex package (v0.3.0)
I would like to lag several columns (e.g. value_1 + value_2 + x - see below), define their number of lags (e.g. 3) and their naming. This is some working tedious/manual code:
library(dplyr)
library(lubridate)
library(data.table)
haves <- data.frame(
id = c("a", "a", "a", "a", "a", "a", "b", "b", "b", "b", "b", "b")
, date = as.Date(c("2015-01-01", "2015-02-01", "2015-03-01", "2015-04-01", "2015-05-01", "2015-06-01", "2015-01-01", "2015-02-01", "2015-03-01", "2015-04-01", "2015-05-01", "2015-06-01"))
, value_1 = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
, value_2 = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
, x = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
)
haves$value_2 <- haves$value_2 + 1
haves$x <- haves$x + 2
haves
wants <- haves %>%
group_by(id) %>%
mutate(
value_1_lag_1 = lag(value_1, n = 1, order_by = date)
, value_1_lag_2 = lag(value_1, n = 2, order_by = date)
, value_1_lag_3 = lag(value_1, n = 3, order_by = date)
, value_2_lag_1 = lag(value_2, n = 1, order_by = date)
, value_2_lag_2 = lag(value_2, n = 2, order_by = date)
, value_2_lag_3 = lag(value_2, n = 3, order_by = date)
, x_lag_1 = lag(x, n = 1, order_by = date)
, x_lag_2 = lag(x, n = 2, order_by = date)
, x_lag_3 = lag(x, n = 3, order_by = date)
)
wants
Someone suggested that this offers a solution, so I tried the data table approach without success:
setDT(haves)
haves[, sapply(1:3, function(x){paste0('', x, '_lag_', 1:3)}) := shift(.SD, 1:3),
by = id, .SDcols = value_1:x][]
It does not produce my wants. This comes closer:
colnames <- colnames(haves)
setDT(haves)
haves[, sapply(1:3, function(x){paste0(colnames[x + 2], x, '_lag_', 1:3)}) := shift(.SD, 1:3), by = id, .SDcols = value_1:x][]
Alternatively, I may just use a loop and a function like this:
appender <- function(df, column, lag){
df %>%
group_by(
id
) %>%
mutate(
!!paste0(column, "_lag_", lag) := lag(!!rlang::sym(column), n = lag, order_by = date)
)
}
temp <- appender(haves, "value_2", 3)
Any help would be very much appreciated. Thanks!
Here's the adapted dplyr solution from the linked answer.
haves %>%
group_by(id) %>%
nest %>%
mutate(data = map(data, ~arrange(., date))) %>%
mutate(lags = map(data, function(dat) {
imap_dfc(dat[-1], ~set_names(map(1:3, lag, x = .x),
paste0(.y, "_lag_", 1:3)))
})) %>%
unnest(c(data, lags))
Is this what you're looking for?
Here is an alternative way through data.table.
library(data.table)
library(lubridate)
#>
#> Attaching package: 'lubridate'
#> The following objects are masked from 'package:data.table':
#>
#> hour, isoweek, mday, minute, month, quarter, second, wday, week,
#> yday, year
#> The following object is masked from 'package:base':
#>
#> date
library(stringr)
haves <- data.frame(
id = c("a", "a", "a", "a", "a", "a", "b", "b", "b", "b", "b", "b")
, date = as.Date(c("2015-01-01", "2015-02-01", "2015-03-01", "2015-04-01", "2015-05-01", "2015-06-01", "2015-01-01", "2015-02-01", "2015-03-01", "2015-04-01", "2015-05-01", "2015-06-01"))
, value_1 = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
, value_2 = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
, value_3 = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
)
setDT(haves)
setorder(haves, date)
N_vars = 3
N_lags = 3
current_vars = paste0( "value_", rep(1:N_vars, each = 1) )
new_vars = paste0( "value_", rep(1:N_vars, each = N_lags), "_lag_", 1:N_lags )
to_define = new_vars[ str_detect(new_vars, "lag_1") ]
haves[ , (new_vars) := shift( .SD, 1:N_lags ), .SDcols = current_vars]
haves
#> id date value_1 value_2 value_3 value_1_lag_1 value_1_lag_2
#> 1: a 2015-01-01 1 1 1 NA NA
#> 2: b 2015-01-01 7 7 7 1 NA
#> 3: a 2015-02-01 2 2 2 7 1
#> 4: b 2015-02-01 8 8 8 2 7
#> 5: a 2015-03-01 3 3 3 8 2
#> 6: b 2015-03-01 9 9 9 3 8
#> 7: a 2015-04-01 4 4 4 9 3
#> 8: b 2015-04-01 10 10 10 4 9
#> 9: a 2015-05-01 5 5 5 10 4
#> 10: b 2015-05-01 11 11 11 5 10
#> 11: a 2015-06-01 6 6 6 11 5
#> 12: b 2015-06-01 12 12 12 6 11
#> value_1_lag_3 value_2_lag_1 value_2_lag_2 value_2_lag_3 value_3_lag_1
#> 1: NA NA NA NA NA
#> 2: NA 1 NA NA 1
#> 3: NA 7 1 NA 7
#> 4: 1 2 7 1 2
#> 5: 7 8 2 7 8
#> 6: 2 3 8 2 3
#> 7: 8 9 3 8 9
#> 8: 3 4 9 3 4
#> 9: 9 10 4 9 10
#> 10: 4 5 10 4 5
#> 11: 10 11 5 10 11
#> 12: 5 6 11 5 6
#> value_3_lag_2 value_3_lag_3
#> 1: NA NA
#> 2: NA NA
#> 3: 1 NA
#> 4: 7 1
#> 5: 2 7
#> 6: 8 2
#> 7: 3 8
#> 8: 9 3
#> 9: 4 9
#> 10: 10 4
#> 11: 5 10
#> 12: 11 5
Created on 2020-05-04 by the reprex package (v0.3.0)
I don't know exactly how to explain it but...
I have a sparse table where each group represents a level. The columns are ordered, it means, the downstream (left) column represents a child node and upstream (right) node represents a parent node.
I'd like a two columns table where the 1st column is the parent node and the 2nd is the child node. If possible, a 3rd columns with the length (sum of the number of final nodes) of the parents.
Follow the example:
>tt <- tibble(
ID = letters[1:8],
`1` = c( 1, 1, 1, 1, 2, 2, 2, 2),
`2` = c( 3, 3, 4, 4, 5, 5, 5, 6),
`3` = c( 7, 7, 8, 9,10,10,11,12)
)
> tt
# A tibble: 8 x 4
ID `1` `2` `3`
<chr> <dbl> <dbl> <dbl>
1 a 1 3 7
2 b 1 3 7
3 c 1 4 8
4 d 1 4 9
5 e 2 5 10
6 f 2 5 10
7 g 2 5 11
8 h 2 6 12
>dput(tt)
structure(list(ID = c("a", "b", "c", "d", "e", "f", "g", "h"),
`1` = c(1, 1, 1, 1, 2, 2, 2, 2), `2` = c(3, 3, 4, 4, 5, 5,
5, 6), `3` = c(7, 7, 8, 9, 10, 10, 11, 12)), row.names = c(NA,
-8L), class = c("tbl_df", "tbl", "data.frame"))
the result should be:
>ttt <- tibble(
parent = c(1,1,2,2,3,4,4, 5, 5, 6, 7,7,8,9,10,10,11,12),
child = c(3,4,5,6,7,8,9,10,11,12, letters[1:8] ),
length = c(4,4,4,4,2,2,2, 3, 3, 1, 2,2,1,1, 2, 2, 1, 1)
)
>ttt
# A tibble: 18 x 3
parent child length
<dbl> <chr> <dbl>
1 1 3 4
2 1 4 4
3 2 5 4
4 2 6 4
5 3 7 2
6 4 8 2
7 4 9 2
8 5 10 3
9 5 11 3
10 6 12 1
11 7 a 2
12 7 b 2
13 8 c 1
14 9 d 1
15 10 e 2
16 10 f 2
17 11 g 1
18 12 h 1
> dput(ttt)
structure(list(parent = c(1, 1, 2, 2, 3, 4, 4, 5, 5, 6, 7, 7,
8, 9, 10, 10, 11, 12), child = c("3", "4", "5", "6", "7", "8",
"9", "10", "11", "12", "a", "b", "c", "d", "e", "f", "g", "h"
), length = c(4, 4, 4, 4, 2, 2, 2, 3, 3, 1, 2, 2, 1, 1, 2, 2,
1, 1)), row.names = c(NA, -18L), class = c("tbl_df", "tbl", "data.frame"
))
Any help is appreciated.
Thanks in advance.
This gets you 90% of the way there:
tt_correct <- tt[, c(2,3,4,1)]
ttt <- do.call(
rbind,
lapply(seq_len(length(tt)-1),
function(i){
DF <- tt_correct[, c(i, i+1)]
names(DF) <- c('parent', 'child')
DF$length <- ave(DF$parent, DF$parent, FUN = length)
unique(DF)
}
)
)
ttt
# A tibble: 18 x 3
parent child length
<dbl> <chr> <dbl>
1 1 3 4
2 1 4 4
3 2 5 4
4 2 6 4
5 3 7 2
6 4 8 2
7 4 9 2
8 5 10 3
9 5 11 3
10 6 12 1
11 7 a 2
12 7 b 2
13 8 c 1
14 9 d 1
15 10 e 2
16 10 f 2
17 11 g 1
18 12 h 1
The first part is correcting the order. Your expected output indicates that the 1st column is a child of the 4th column. The lapply() statement largely walks along the data.frame and stacks the data.
This is 90% of the way because the answer doesn't agree with your expected output for lengths. I think this is correct but I could be wrong.
Finally, and I'm not that good with igraph, you could likely find additional information doing:
library(igraph)
plot(graph_from_data_frame(ttt[, 1:2]))
I am trying to add a column to my existing data set.
The data set has three columns:
Student (which is the column with the participant ID),
Week (the number of the week of the year during which the data were collected),
and
Day (the number of the weekday during which the data were
collected).
Now, a new column Obs that I am trying to create would contain a progressive number (from 1 to n) referring to the week during which every student was tested.
I have tried to use group_by in combination with rep but it does not seem to produce the result I want:
Week <- c(1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 4)
Day <- c(1, 2, 3, 2, 3, 5, 1, 3, 2, 3, 4, 5)
Student <- c("A", "A", "A", "B", "B", "B", "B", "B", "C", "C", "C", "C")
fake.db <- data.frame(Student, Week, Day)
library(dplyr)
fake.db %>%
group_by(Student) %>%
mutate(Obs = rep(1:length(Student), each = Week))
# Student Week Day Obs
# <fct> <dbl> <dbl> <int>
# 1 A 1 1 1
# 2 A 1 2 2
# 3 A 1 3 3
# 4 B 2 2 1
# 5 B 2 3 2
# 6 B 2 5 3
# 7 B 3 1 4
# 8 B 3 3 5
# 9 C 4 2 1
#10 C 4 3 2
#11 C 4 4 3
#12 C 4 5 4
What I would like to obtain is different. For the first week of data collection, 1 should be reported, and for the students for whom data were collected during a second week, 2 should be reported, etc.:
# Student Week Day Obs
#1 A 1 1 1
#2 A 1 2 1
#3 A 1 3 1
#4 B 2 2 1
#5 B 2 3 1
#6 B 2 5 1
#7 B 3 1 2
#8 B 3 3 2
#9 C 4 2 1
#10 C 4 3 1
#11 C 4 4 1
#12 C 4 5 1
One dplyr possibility could be:
fake.db %>%
group_by(Student) %>%
mutate(Obs = cumsum(!duplicated(Week)))
Student Week Day Obs
<fct> <dbl> <dbl> <int>
1 A 1 1 1
2 A 1 2 1
3 A 1 3 1
4 B 2 2 1
5 B 2 3 1
6 B 2 5 1
7 B 3 1 2
8 B 3 3 2
9 C 4 2 1
10 C 4 3 1
11 C 4 4 1
12 C 4 5 1
It groups by "Student" column and calculates the cumulative sum of non-duplicate "Week" values.
Or:
fake.db %>%
group_by(Student) %>%
mutate(Obs = with(rle(Week), rep(seq_along(lengths), lengths)))
It groups by "Student" column and creates a run-length type group ID around "Week" column".
Or:
fake.db %>%
group_by(Student) %>%
mutate(Obs = dense_rank(Week))
It groups by "Student" column and ranks the values in "Week" column.
What I understand the issue to be is that you want to count the weeks since the first test week for each student. I.e. Week 2 is student B's first week of testing, so it gets Obs = 1. That means you can do a grouped mutate:
library(dplyr)
fake.db <- structure(list(Student = structure(c(1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L), .Label = c("A", "B", "C"), class = "factor"), Week = c(1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 4), Day = c(1, 2, 3, 2, 3, 5, 1, 3, 2, 3, 4, 5)), class = "data.frame", row.names = c(NA, -12L))
fake.db %>%
group_by(Student) %>%
mutate(Obs = Week - min(Week) + 1)
#> # A tibble: 12 x 4
#> # Groups: Student [3]
#> Student Week Day Obs
#> <fct> <dbl> <dbl> <dbl>
#> 1 A 1 1 1
#> 2 A 1 2 1
#> 3 A 1 3 1
#> 4 B 2 2 1
#> 5 B 2 3 1
#> 6 B 2 5 1
#> 7 B 3 1 2
#> 8 B 3 3 2
#> 9 C 4 2 1
#> 10 C 4 3 1
#> 11 C 4 4 1
#> 12 C 4 5 1
Created on 2019-05-10 by the reprex package (v0.2.1)
A brief method with by
unlist(by(fake.db, fake.db[, 1], function(x) as.numeric(factor(x[, 2]))))
# A1 A2 A3 B1 B2 B3 B4 B5 C1 C2 C3 C4
# 1 1 1 1 1 1 2 2 1 1 1 1
Data
fake.db <- structure(list(Student = structure(c(1L, 1L, 1L, 2L, 2L, 2L,
2L, 2L, 3L, 3L, 3L, 3L), .Label = c("A", "B", "C"), class = "factor"),
Week = c(1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 4), Day = c(1,
2, 3, 2, 3, 5, 1, 3, 2, 3, 4, 5)), class = "data.frame", row.names = c(NA,
-12L))
You can see if there is a non-zero difference
fake.db %>%
group_by(Student) %>%
arrange(Week) %>%
mutate(Obs = cumsum(c(1, diff(Week)!=0)))
or if they values arne't numeric, you can compare to the lag value
fake.db %>%
group_by(Student) %>%
arrange(Week) %>%
mutate(Obs = cumsum(Week != lag(Week, default=first(Week))) + 1)
I have a dataset of people (pid) of different types (type2=c("dad", "mom", "kid"; and for ease, type=c("a", "b", "c")) nested in households (hid) with repeated measurements (time).
Some variables like v1_ are asked to everyone, but the values are spread across three columns. For instance, v1_a contains the values for all of the dads (type==a).
Variables like v2_ are only asked of dads and moms (a's and b's), and the values are spread across two columns.
Variables like v3 are also only asked to dads and moms, but the values are contained in one column.
Variables like v4 are asked to everyone, and the values are contained in one column.
Have:
hid pid type type2 time v1_a v1_b v1_c v2_a v2_b v3 v4
1 1 1 a dad 1 6 NA NA 2 NA 4 3
2 1 2 b mom 1 NA 2 NA NA 5 6 6
3 1 3 c kid 1 NA NA 1 NA NA NA 5
4 2 4 a dad 1 3 NA NA 6 NA 2 6
5 2 5 b mom 1 NA 5 NA NA 2 4 3
6 2 6 c kid 1 NA NA 3 NA NA NA 5
7 1 1 a dad 2 3 NA NA 2 NA 4 3
8 1 2 b mom 2 NA 3 NA NA 5 6 6
9 1 3 c kid 2 NA NA 2 NA NA NA 5
10 2 4 a dad 2 2 NA NA 6 NA 2 6
11 2 5 b mom 2 NA 3 NA NA 2 4 3
12 2 6 c kid 2 NA NA 2 NA NA NA 5
Here is the end result I want:
hid pid type type2 time v1 v2 v3 v4
1 1 1 a dad 1 6 2 4 3
2 1 2 b mom 1 2 5 6 6
3 1 3 c kid 1 1 NA NA 5
4 2 4 a dad 1 3 6 2 6
5 2 5 b mom 1 5 2 4 3
6 2 6 c kid 1 3 NA NA 5
7 1 1 a dad 2 3 2 4 3
8 1 2 b mom 2 3 5 6 6
9 1 3 c kid 2 2 NA NA 5
10 2 4 a dad 2 2 6 2 6
11 2 5 b mom 2 3 2 4 3
12 2 6 c kid 2 2 NA NA 5
I'm looking for a tidyverse approach that will handle a larger actual use case of mixed variables as shown here. The variable naming is consistent. Where do I go after gather()?
library(tidyverse)
df_have <- data.frame(hid=c(1, 1, 1, 2, 2, 2,
1, 1, 1, 2, 2, 2),
pid=c(1, 2, 3, 4, 5, 6,
1, 2, 3, 4, 5, 6),
type=c("a", "b", "c", "a", "b", "c",
"a", "b", "c", "a", "b", "c"),
type2=c("dad", "mom", "kid", "dad", "mom", "kid",
"dad", "mom", "kid", "dad", "mom", "kid"),
time=c(1, 1, 1, 1, 1, 1,
2, 2, 2, 2, 2, 2),
v1_a=c(6, NA, NA, 3, NA, NA,
3, NA, NA, 2, NA, NA),
v1_b=c(NA, 2, NA, NA, 5, NA,
NA, 3, NA, NA, 3, NA),
v1_c=c(NA, NA, 1, NA, NA, 3,
NA, NA, 2, NA, NA, 2),
v2_a=c(2, NA, NA, 6, NA, NA,
2, NA, NA, 6, NA, NA),
v2_b=c(NA, 5, NA, NA, 2, NA,
NA, 5, NA, NA, 2, NA),
v3=c(4, 6, NA, 2, 4, NA,
4, 6, NA, 2, 4, NA),
v4=c(3, 6, 5, 6, 3, 5,
3, 6, 5, 6, 3, 5)
)
df_want <- data.frame(hid=c(1, 1, 1, 2, 2, 2,
1, 1, 1, 2, 2, 2),
pid=c(1, 2, 3, 4, 5, 6,
1, 2, 3, 4, 5, 6),
type=c("a", "b", "c", "a", "b", "c",
"a", "b", "c", "a", "b", "c"),
type2=c("dad", "mom", "kid", "dad", "mom", "kid",
"dad", "mom", "kid", "dad", "mom", "kid"),
time=c(1, 1, 1, 1, 1, 1,
2, 2, 2, 2, 2, 2),
v1=c(6, 2, 1, 3, 5, 3,
3, 3, 2, 2, 3, 2),
v2=c(2, 5, NA, 6, 2, NA,
2, 5, NA, 6, 2, NA),
v3=c(4, 6, NA, 2, 4, NA,
4, 6, NA, 2, 4, NA),
v4=c(3, 6, 5, 6, 3, 5,
3, 6, 5, 6, 3, 5)
)
df_have %>%
gather(key, value, -hid, -pid, -type, -type2, -time)
Here is another idea using coalesce from dplyr and map from purrr.
library(tidyverse)
# Set target column names
cols <- paste0("v", 1:4)
# Coalesce the numbers based on column names
nums <- map(cols, ~coalesce(!!!as.list(df_have %>% select(starts_with(.x)))))
# Create a data frame
nums_df <- nums %>%
setNames(cols) %>%
as_data_frame()
# Create the final output by bind_cols
df_test <- df_have %>%
select(-starts_with("v")) %>%
bind_cols(nums_df)
df_test
# hid pid type type2 time v1 v2 v3 v4
# 1 1 1 a dad 1 6 2 4 3
# 2 1 2 b mom 1 2 5 6 6
# 3 1 3 c kid 1 1 NA NA 5
# 4 2 4 a dad 1 3 6 2 6
# 5 2 5 b mom 1 5 2 4 3
# 6 2 6 c kid 1 3 NA NA 5
# 7 1 1 a dad 2 3 2 4 3
# 8 1 2 b mom 2 3 5 6 6
# 9 1 3 c kid 2 2 NA NA 5
# 10 2 4 a dad 2 2 6 2 6
# 11 2 5 b mom 2 3 2 4 3
# 12 2 6 c kid 2 2 NA NA 5
This gets me there, but the filter(!is.na(value)) step seems like a hack. Better ideas?
df_test <-
df_have %>%
gather(key, value, -hid, -pid, -type, -time, -type2) %>%
mutate(key = str_replace(key, "_.*", "")) %>%
filter(!is.na(value)) %>%
spread(key, value) %>%
arrange(time, hid, type, pid)
Update from #www:
df_test <-
df_have %>%
gather(key, value, -hid, -pid, -type, -time, -type2, na.rm=TRUE) %>%
mutate(key = str_replace(key, "_.*", "")) %>%
spread(key, value) %>%
arrange(time, hid, type, pid)