replacing NAs with preceding strings in R [duplicate] - r

I have a dataset looking like this
df <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(NA, NA, 12, 13, NA, 5, NA, NA, NA, 1))
I want an output like this, so that last observations are carried forward (by group) unless there are only NA values before one fillied-in value then I want last-observation carried backward:
df <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(12, 12, 12, 13, 13, 5, 5, 5, 1, 1))
I have been working with dplyr and na.locf from the zoo package. SO far my approach has been this:
df%>%
group_by(PID%>%
mutate_all(funs(na.locf(., na.rm = FALSE)))
However, this only does last observation carried forward. The specification "fromLast" in the na.locf function does last observation carried backward.
But how do I connect these two, so that both functions are used:
na.LOCF if there are no NA values before the first filled-in value
na.LOCF(fromLast) meaning last observation carried backward if there are NA values before the first value that is filled-in.
Thank you so much in advance!

This should work :
library(tidyverse)
df <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(NA, NA, 12, 13, NA, 5, NA, NA, NA, 1))
df2 <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(12, 12, 12, 13, 13, 5, 5, 5, 1, 1))
df <- df %>%
group_by(ID) %>%
fill(values, .direction = "downup") %>%
fill(values, .direction = "updown")

Related

Using map() function to apply for each element

I need, with the help of the map() function, apply the above for each element
How can I do so?
As dt is of class data.table, you can make a vector of columns of interest (i.e. your items; below I use grepl on the names), and then apply your weighting function to each of those columns using .SD and .SDcols, with by
qs = names(dt)[grepl("^q", names(dt))]
dt[, (paste0(qs,"wt")):=lapply(.SD, \(q) 1/(sum(!is.na(q))/.N)),
.(sex, education_code, age), .SDcols = qs]
As mentioned in the comments, you miss a dt <- in your dt[, .(ID, education_code, age, sex, item = q1_1)] which makes the column item unavailable in the following line dt[, no_respond := is.na(item)].
Your weighting scheme is not absolutely clear to me however, assuming you want to do what is done in your code here, I would go with dplyr solution to iterate over columns.
# your data without no_respond column and correcting missing value in q2_3
dt <- data.table::data.table(
ID = c(1,2,3,4, 5, 6, 7, 8, 9, 10),
education_code = c(20,50,20,60, 20, 10,5, 12, 12, 12),
age = c(87,67,56,52, 34, 56, 67, 78, 23, 34),
sex = c("F","M","M","M", "F","M","M","M", "M","M"),
q1_1 = c(NA,1,5,3, 1, NA, 3, 4, 5,1),
q1_2 = c(NA,1,5,3, 1, 2, NA, 4, 5,1),
q1_3 = c(NA,1,5,3, 1, 2, 3, 4, 5,1),
q1_text = c(NA,1,5,3, 1, 2, 3, 4, 5,1),
q2_1 = c(NA,1,5,3, 1, 2, 3, 4, 5,1),
q2_2 = c(NA,1,5,3, 1, 2, 3, 4, 5,1),
q2_3 = c(NA,1,5,3, 1, NA, NA, 4, 5,1),
q2_text = c(NA,1,5,3, 1, NA, 3, 4, 5,1))
dt %>%
group_by(sex, education_code, age) %>% #groups the df by sex, education_code, age
add_count() %>% #add a column with number of rows in each group
mutate(across(starts_with("q"), #for each column starting with "q"
~ 1/(sum(!is.na(.))/n), #create a new column following your weight calculation
.names = '{.col}_wgt')) %>% #naming the new column with suffix "_wgt" to original name
ungroup()

Last observation carried forward and last observation carried backward in R

I have a dataset looking like this
df <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(NA, NA, 12, 13, NA, 5, NA, NA, NA, 1))
I want an output like this, so that last observations are carried forward (by group) unless there are only NA values before one fillied-in value then I want last-observation carried backward:
df <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(12, 12, 12, 13, 13, 5, 5, 5, 1, 1))
I have been working with dplyr and na.locf from the zoo package. SO far my approach has been this:
df%>%
group_by(PID%>%
mutate_all(funs(na.locf(., na.rm = FALSE)))
However, this only does last observation carried forward. The specification "fromLast" in the na.locf function does last observation carried backward.
But how do I connect these two, so that both functions are used:
na.LOCF if there are no NA values before the first filled-in value
na.LOCF(fromLast) meaning last observation carried backward if there are NA values before the first value that is filled-in.
Thank you so much in advance!
This should work :
library(tidyverse)
df <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(NA, NA, 12, 13, NA, 5, NA, NA, NA, 1))
df2 <- data.frame(ID=c(1,1,1,1,1,2,2,2,3,3), values=c(12, 12, 12, 13, 13, 5, 5, 5, 1, 1))
df <- df %>%
group_by(ID) %>%
fill(values, .direction = "downup") %>%
fill(values, .direction = "updown")

Returning values from a column based on the last value of another column

I have a dataset like this:
data <- data.frame(Time = c(1,4,6,9,11,13,16, 25, 32, 65),
A = c(10, NA, 13, 2, 32, 19, 32, 34, 93, 12),
B = c(1, 99, 32, 31, 12, 13, NA, 13, NA, NA),
C = c(2, 32, NA, NA, NA, NA, NA, NA, NA, NA))
What I want to retrieve are the values in Time that corresponds to the last numerical value in A, B, and C.
For example, the last numerical values for A, B, and C are 12, 13, and 32 respectively.
So, the Time values that correspond are 65, 25, and 4.
I've tried something like data[which(data$Time== max(data$A)), ], but this doesn't work.
We can multiply the row index with the logical matrix, and get the colMaxs (from matrixStats) to subset the 'Time' column
library(matrixStats)
data$Time[colMaxs((!is.na(data[-1])) * row(data[-1]))]
#[1] 65 25 4
Or using base R, we get the index with which/arr.ind, get the max index using a group by operation (tapply) and use that to extract the 'Time' value
m1 <- which(!is.na(data[-1]), arr.ind = TRUE)
data$Time[tapply(m1[,1], m1[,2], FUN = max)]
#[1] 65 25 4
Or with summarise/across in the devel version of dplyr
library(dplyr)
data %>%
summarise(across(A:C, ~ tail(Time[!is.na(.)], 1)))
# A B C
#1 65 25 4
Or using summarise_at with the current version of dplyr
data %>%
summarise_at(vars(A:C), ~ tail(Time[!is.na(.)], 1))

How can I estimate a function in a group?

I have a data frame with 1530 obs of 6 varaibles. In this dataframe there 51 assets with 30 obs each. I tried to apply de MACD function to obtain two values: macd and signal but show up an error. This is an example:
macdusdt <- filtusdt %>% group_by(symbol) %>% do(tail(., n = 30))
macd1m <- macdusdt %>%
mutate (signals = MACD(macdusdt$lastPrice,
nFast = 12, nSlow = 26, nSig = 9, maType = "EMA", percent = T))
Error: Column signals must be length 30 (the group size) or one, not 3060
I want to apply de MACD function to every asset in the data frame. The database is here: https://www.dropbox.com/s/ww8stgsspqi8tef/macdusdt.xlsx?dl=0
Based on the data provided, it is giving an error when applied the code
Error in EMA(c(NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, :
n > number of non-NA values in column(s) 1
To prevent that we can do
library(dplyr)
library(TTR)
filtusdt %>%
group_by(symbol) %>%
slice(tail(row_number(), 30)) %>%
mutate(signals = if(n() < sum(is.na(lastPrice))) MACD(lastPrice,
nFast = 12, nSlow = 26, nSig = 9, maType = "EMA", percent = TRUE) else NA)
It could be an issue because of the subset dataset provided

Replace NAs in one vector with sequential elements of another vector

I'd like to replace NA elements of a vector with elements from a sequence, for example:
x <- c(1, NA, 5, NA, NA, 2, 12, NA)
replace.seq <- -1:-4 # Can assume length(replace.seq) == sum(is.na(x))
goal <- c(1, -1, 5, -2, -3, 2, 12, -4)
What's an efficient way to do this? I'd prefer to avoid sorting x.
Per #akrun:
x[is.na(x)] <- replace.seq
You can use replace:
x <- replace(x, is.na(x), replace.seq)

Resources