Daily to Weekly and Daily to Monthly Aggregation

[taranpreetk]

Salutation for the day.

Thank you for the fantastic package design.

I was unsure if this package could aggregate time series from daily to weekly and daily to monthly because when I attempt to do so, I receive an error saying there isn't an option for "weekly" or "monthly".

I have put a reproducible example for this.

`
library(Sim.DiffProc)
library(stats)
library(tempdisagg)
set.seed(1234)

dates <- seq(as.Date("2019-01-01"), as.Date("2023-12-31"), by = "1 day")

X <- GBM(N = length(dates)-1, M = 1, x0 = 50, y= 200)
daily_data <- data.frame(dates,X)
daily_data <- ts(daily_data$X1, start = c(2019,1,1), end = c(2022,12,31), frequency = 365)
weekly_data <- ta(daily_data, conversion = "sum", to = "weekly")
monthly_data <- ta(daily_data, conversion = "sum", to = "monthly")
`
Error in Console:

I'm not sure how to get this working as it is needed for the use case I'm working on. Any help on this will be appreciated.

Thanks in advance.

[christophsax]

The main point of tempdisagg, is to disaggregate:

• The classic methods allow you to convert annual or quarterly series into regular quarterly or monthly series. See here.

• tempdisagg now allows disaggregating into daily series.
  See the last example in ?td.

• With some trickery, you can do quarterly to weekly in a way that takes into account that different overlap between weeks and quarters in different years. See here

• Finally, you can convert to hourly data.
  See here

What you are up to it to aggregate, which is a far easier task. tsbox can solve the problem you are struggling with:

library(dplyr)
library(tsbox)

dates = seq(as.Date("2019-01-01"), as.Date("2023-12-31"), by = "1 day")
daily <- tibble(
  time = dates,
  value = rnorm(length(dates))
) 

daily
#> # A tibble: 1,826 × 2
#>    time         value
#>    <date>       <dbl>
#>  1 2019-01-01  0.350 
#>  2 2019-01-02 -1.12  
#>  3 2019-01-03  0.384 
#>  4 2019-01-04 -0.0618
#>  5 2019-01-05  0.764 
#>  6 2019-01-06  1.09  
#>  7 2019-01-07  0.660 
#>  8 2019-01-08 -0.411 
#>  9 2019-01-09  0.893 
#> 10 2019-01-10  0.0300
#> # … with 1,816 more rows

tsbox::ts_frequency(daily, to = "month")
#> # A tibble: 60 × 2
#>    time          value
#>    <date>        <dbl>
#>  1 2019-01-01  0.111  
#>  2 2019-02-01  0.183  
#>  3 2019-03-01 -0.108  
#>  4 2019-04-01 -0.103  
#>  5 2019-05-01  0.00898
#>  6 2019-06-01  0.111  
#>  7 2019-07-01 -0.141  
#>  8 2019-08-01  0.213  
#>  9 2019-09-01 -0.100  
#> 10 2019-10-01  0.134  
#> # … with 50 more rows
tsbox::ts_frequency(daily, to = "year")
#> # A tibble: 5 × 2
#>   time          value
#>   <date>        <dbl>
#> 1 2019-01-01  0.0375 
#> 2 2020-01-01  0.00706
#> 3 2021-01-01  0.0208 
#> 4 2022-01-01 -0.00932
#> 5 2023-01-01  0.0706
tsbox::ts_frequency(daily, to = "week")
#> # A tibble: 260 × 2
#>    time         value
#>    <date>       <dbl>
#>  1 2019-01-06  0.342 
#>  2 2019-01-13  0.313 
#>  3 2019-01-20 -0.226 
#>  4 2019-01-27 -0.108 
#>  5 2019-02-03 -0.292 
#>  6 2019-02-10  0.292 
#>  7 2019-02-17  0.658 
#>  8 2019-02-24 -0.392 
#>  9 2019-03-03  0.0588
#> 10 2019-03-10 -0.581 
#> # … with 250 more rows

And if you need seem instead of averages:

tsbox::ts_frequency(daily, to = "week", aggregate = "sum")
#> # A tibble: 260 × 2
#>    time        value
#>    <date>      <dbl>
#>  1 2019-01-06  2.40 
#>  2 2019-01-13  2.19 
#>  3 2019-01-20 -1.59 
#>  4 2019-01-27 -0.759
#>  5 2019-02-03 -2.05 
#>  6 2019-02-10  2.04 
#>  7 2019-02-17  4.61 
#>  8 2019-02-24 -2.74 
#>  9 2019-03-03  0.412
#> 10 2019-03-10 -4.06 
#> # … with 250 more rows

^{Created on 2023-04-20 with reprex v2.0.2}