Custom time periods

[ZherlockHolm]

Hello community,

I am working with a dataset with about 69,000 observations over a 12-year period. Observations are linked to a date-variable (format: "2012-10-27"). I am trying to make a new column where observations made between the 18th of the first month and the 17th of the following month are lumped together (i.e. where "2012-10-27", "2012-11-04" & "2012-11-14" all would be labelled as "2012-OctNov" and "2012-11-18" & "2012-12-15" belong to "2012-NovDec" etc.) for all 12 years coming out to 12*12 = 144 labels I guess.

I have googled and looked through previous questions, but I can't seem to find any leads. I'm not sure which categories this question falls under, but I hope someone can help!

[andresrcs]

You can create a custom indexing function to use with tsibble package, see this related thread:

Aggregating Time Series tsibbles on non-calendar years

Lubridate functions are mostly built for calendar year. For this case, you need a custom function that does different binning. An Aussie version of financial_year() is defined for this example: library(tsibble) library(dplyr, warn.conflicts = FALSE) library(lubridate, warn.conflicts = FALSE) q_data <- tsibbledata::gafa_stock %>% index_by(yq0 = yearquarter(Date)) %>% summarize(close = last(Close)) financial_year <- function(date) { year <- year(date) ifelse(quarter(date) <= 2, year, year + 1) } q_data %>% index_by(y0 = ~ financial_year(.)) #> # A tsibble: 20 x 3 [1Q] #> # Groups: @ y0 [6] #> yq0 close y0 #> <qtr> <dbl> <dbl> #> 1 2014 Q1 554. 2014 #> 2 20…

Not simple solution comes to my mind at this moment for the custom function, but I have a hunch that there must be one. If my brain wakes up later and I figure it out I will come back with more specific help.