CODEWITHSUNDEEP
rtime-seriesxts
Yun Ching
Yun Ching

Reputation: 675

Updating time series with another time series

Suppose I have two time series

(x <- xts(c(5,4:10), Sys.Date()+3:10))
(y <- xts(1:6, Sys.Date()+1:6))

merge(x,y)

            x  y
2018-04-20 NA  1
2018-04-21 NA  2
2018-04-22  5  3
2018-04-23  4  4
2018-04-24  5  5
2018-04-25  6  6
2018-04-26  7 NA
2018-04-27  8 NA
2018-04-28  9 NA
2018-04-29 10 NA

How do I get a time series that takes the value of y if it is present, but falls back to x only if y does not have a value at a particular date? For example, I want the values in z as below:

            x  y z
2018-04-20 NA  1 1
2018-04-21 NA  2 2 
2018-04-22  5  3 3 
2018-04-23  4  4 4 
2018-04-24  5  5 5
2018-04-25  6  6 6
2018-04-26  7 NA 7 
2018-04-27  8 NA 8 
2018-04-28  9 NA 9
2018-04-29 10 NA 10

Upvotes: 1

Views: 130

Answers (2)

Joshua Ulrich
Joshua Ulrich

Reputation: 176688

Here's another solution, and it does not use ifelse().

# Merge the two objects first
z <- merge(x,y)

# Create a new column from a copy of 'y'
z$z <- z$y

# Missing values in column 'z'
zNA <- is.na(z$z)

# Fill all missing values in column 'z' with values from column 'x'
z[zNA, "z"] <- z[zNA, "x"]

I personally avoid ifelse() because the object it returns is not always intuitively predictable.

Upvotes: 2

maraboule
maraboule

Reputation: 363

Would that solve your problem ?

z <- rbind(x,y)
z <- z[!duplicated(z)]

Edit:

Note that rbind() binds xts objects by row in order of priority, i.e. the rows of y will be before rows of x because y was put before x in the function call 

> rbind(y, x) 
           [,1] 
2018-04-21    1 
2018-04-22    2 
2018-04-23    3 
2018-04-23    5 
2018-04-24    4 
2018-04-24    4 
2018-04-25    5 
2018-04-25    5 
2018-04-26    6 
2018-04-26    6 
2018-04-27    7 
2018-04-28    8 
2018-04-29    9 
2018-04-30   10

Then I noticed that z[!duplicated(z)] is not correct actually. The duplicates you want to take out are duplicates in index so try that out. 

z <- rbind(y, x) 
z <- z[!duplicated(index(z))]

I am a fervent user of xts and I always try to use its functions because they usually do all the work for you in a very readable and natural fashion. But when I can't find my way out of it I often use the powerful ifelse(), prince of efficiency and destroyer of ambiguity. 

z <- merge(x, y) 
zz <- ifelse(is.na(z[, 2]), z[, 1], z[, 2])

Now zz is not an xts object anymore so have to do

zz <- xts(zz, index(z))

But you will be left with duplicates. Since I don't know from your answer if you actually want to keep them, I'll leave you the choice.

Hope that helps :)

Upvotes: 2

Related Questions