In mitsuoxv/tqr: Add tq_xx functions to tidyquant package to facilitate economic analyses

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

Updated: r Sys.Date()

tqr: add-on to tsibble, inspired by tidyquant

When I utilize tidyquant package for economic analyses, I often find myself writing similar functions repeatedly. I also find the idea of time aware dataframe of tsibble package may help me deal with time series data more confidently. So I have decided to build this package to facilitate my work.

As of version 0.0.0.9011, I have added time-wise functions to transform a numeric vector, and a function factory using those time-wise functions, rewritten functions to transform a tsibble using that function factory, and deprecated old function factories.

1. Time-wise functions to transform a numeric vector
2. moving_average: transform into moving averages
3. growth_rate: transform into growth rates
4. season_adjust: transform into seasonally adjusted values
5. A function factory to produce functions to transform a tsibble
6. cal_time_wise: produce functions using time_wise vector functions
7. Functions to transform a tsibble
8. tq_diff: transform numeric columns into differences
9. tq_ma: transform numeric columns into moving averages
10. tq_gr: transform numeric columns into growth rates
11. tq_sa: transform numeric columns into seasonally adjusted values
12. Deprecated function factories
13. cal_factory: produce functions
14. cal_factory_zoo: produce functions utilizing zoo package
15. cal_factory_ts: produce functions utilizing ts class
16. cal_factory_xts: produce functions utilizing xts package

Installation

This package is only for my own use for now. "You" means "future me".

You can install the development version with:

remotes::install_github("mitsuoxv/tqr")

Libraries

As I read Hyndman, R.J., & Athanasopoulos, G. (2021) Forecasting: principles and practice, 3rd edition, OTexts: Melbourne, Australia. OTexts.com/fpp3., I use fpp3 meta package here.

library(fpp3)
library(tqr)

Transform a tsibble into lower frequency

If you are familiar with tsibble functions, like index_by and group_by_key, you can convert to lower frequency.

aus_livestock

aus_livestock %>% 
  index_by(Quarter = yearquarter(Month)) %>% 
  group_by_key() %>% 
  summarize(Count = sum(Count))

If you are familiar with dplyr and tidyselect functions, you can convert multiple variables at once.

aus_production

aus_production %>% 
  index_by(Year = year(Quarter)) %>% 
  summarize(across(!Quarter, sum))

Time-wise functions to transform a numeric vector are useful and flexible

In the above examples, sum function transforms a numeric vector.

tsibble package provides difference function, which also transforms a time-wise numeric vector. Using it, you can create a new variable "diff".

aus_arrivals %>% 
  group_by_key() %>% 
  mutate(diff = difference(Arrivals)) %>% 
  ungroup()

Or you can transform a existing variable "Arrivals". You can use difference function flexibly.

aus_arrivals %>% 
  group_by_key() %>% 
  mutate(Arrivals = difference(Arrivals)) %>% 
  ungroup()

cal_time_wise: function factory to produce a function to transform a tsibble

cal_time_wise function is functionized from the last example, and is a function factory. It receives a time-wise function to transform a numeric vector as its first argument, and it returns a function to transform a tsibble.

tq_diff: calculate differences

tq_diff function is created like below. tq_ma, tq_gr and tq_sa functions are also created by this function factory.

tq_diff <- cal_time_wise(
  tsibble::difference
)

The variables, which these created functions transform, are all numeric variables which are not key or index variables.

In the below example, "Month" is the index variable, and "State" and "Industry" are the key variables. "Series ID" is a character vector. So tq_diff transforms just "Turnover", a numeric variable.

aus_retail %>% 
  tq_diff()

For arguments other than the first one, a tsibble, of tq_diff function, look at tsibble::difference document. Note that n in the previous version of tq_diff is changed to lag.

"order_by" argument in time-wise functions to transform a numeric vector

You can find "order_by" argument in tsibble::difference document. It is optional in difference function.

tsbl <- tsibble(year = 2000:2005, value = (0:5)^2, index = year)
scrambled <- tsbl %>% slice(sample(nrow(tsbl)))

wrong <- mutate(scrambled, diff = difference(value))
arrange(wrong, year)

right <- mutate(scrambled, diff = difference(value, order_by = year))
arrange(right, year)

cal_time_wise function assigns the index variable in a tsibble to "order_by" argument of difference function. As a result, tq_diff function returns right even if the rows are reshuffled.

tsbl %>% 
  tq_diff()

scrambled %>% 
  tq_diff() %>% 
  arrange(year)

moving_average and tq_ma: calculate moving averages

moving_average function calculates moving averages using slider package. "n" specifies the window width. "order_by" is optional, and works in the same way as in difference function. For other arguments, refer to its document.

aus_livestock %>% 
  mutate(Count = if_else(row_number() == 4, NA_real_, Count)) %>% 
  group_by_key() %>% 
  mutate(
    ma3 = moving_average(Count, n = 3),
    ma3_na.rm = moving_average(Count, n = 3, na.rm = TRUE),
    ma3_left = moving_average(Count, n = 3, .align = "left")
    ) %>% 
  ungroup()

tq_ma function transforms "Count" variable. For arguments other than the first one, a tsibble, look at moving_average document.

aus_livestock %>% 
  tq_ma(n = 3)

growth_rate and tq_gr: calculate growth rates

growth_rate function calculates growth rates. "n" specifies the lag. "order_by" is optional, and works in the same way as in difference function. For other arguments, refer to its document.

aus_arrivals %>% 
  group_by_key() %>% 
  mutate(
    gr_yoy = growth_rate(Arrivals, n = 4),
    gr_annualized = growth_rate(Arrivals, n = 1, annualize = 4),
    gr_yoy_not_pct = growth_rate(Arrivals, n = 4, pct = FALSE)
    )

tq_gr function transforms "Arrivals" variable in this example. For arguments other than the first one, a tsibble, look at growth_rate document.

aus_arrivals %>% 
  tq_gr(n = 4)

season_adjust and tq_sa: calculate seasonally adjusted values

season_adjust function calculates seasonally adjusted values using season package. It requires not just "x" argument, a numeric vector, but also "order_by" argument, a time vector. For other arguments, refer to season document.

aus_arrivals %>% 
  group_by_key() %>% 
  mutate(sa = season_adjust(Arrivals, Quarter)) %>% 
  ungroup()

tq_sa function transforms "Arrivals" variable in this example.

aus_arrivals %>% 
  tq_sa()

tq_sa function returns right even if the rows are reshuffled, as tq_diff, tq_ma and tq_gr functions do.

aus_arrivals %>% 
  slice(sample(nrow(aus_arrivals))) %>% 
  tq_sa() %>% 
  arrange(Origin, Quarter)

EOL

mitsuoxv/tqr documentation built on Nov. 14, 2021, 7:18 p.m.