How do I fill in Dates in ts() function, when there is missing data for many months for over 2 years?

Question

Here is the data:

e <- data.frame( date = c("2016-03-08", "2016-05-19" ,"2016-05-19" ,"2016-09-02" ,"2016-09-02", "2016-11-23", "2016-12-29","2017-02-08" ,"2017-07-24", "2017-07-26" ,"2018-04-05" ,"2018-06-01", "2019-02-07" ,"2019-03-25"), price = c(1300, 1300, 1300 ,1300 ,1300 ,1300 ,1300 ,1300 ,1300 ,1300 ,1375 ,1375 ,1405 ,1405))

summary(e)

When I use ts(), I am not sure why the values are changing. Could someone explain?

b <- ts(e$price, start = c(2016,3), end = c(2019,3), frequency = 12)

Output: Mar 2013 has the right price that is 1300 , but Mar 2019 has the wrong value, it has 1300 instead of 1405

How to fix this?

Answer 1

Your data does not represent a time series since mathematically a time series is a function from unique times to values. Also ts does not handle daily data well.

If we omit all but the last data point in any year/month then we can convert it to a monthly ts series by reading it into a zoo series and in doing so converting the date column to yearmon class and aggregating on identical year/months using the last value. Then convert that to ts class.

library(zoo)

z <- read.zoo(e, FUN = as.yearmon, aggregate = function(x) tail(x, 1))
as.ts(z)

giving:

      Jan  Feb  Mar  Apr  May  Jun  Jul  Aug  Sep  Oct  Nov  Dec
2016           1300   NA 1300   NA   NA   NA 1300   NA 1300 1300
2017   NA 1300   NA   NA   NA   NA 1300   NA   NA   NA   NA   NA
2018   NA   NA   NA 1375   NA 1375   NA   NA   NA   NA   NA   NA
2019   NA 1405 1405