,
Cathy Chamberlin [rev],
Nunes Matt [rev]| Title: | Class-Agnostic Time Series |
|---|---|
| Description: | Time series toolkit with identical behavior for all time series classes: 'ts','xts', 'data.frame', 'data.table', 'tibble', 'zoo', 'timeSeries', 'tsibble', 'tis' or 'irts'. Also converts reliably between these classes. |
| Authors: | Christoph Sax [aut, cre] ,
Cathy Chamberlin [rev],
Nunes Matt [rev] |
| Maintainer: | Christoph Sax <[email protected]> |
| License: | GPL-3 |
| Version: | 0.4.2 |
| Built: | 2024-12-19 02:51:59 UTC |
| Source: | https://github.com/ropensci/tsbox |
The R ecosystem knows a vast number of time series classes: ts, xts, zoo, tsibble, tibbletime, tis, or timeSeries. The plethora of standards causes confusion. As different packages rely on different classes, it is hard to use them in the same analysis. tsbox provides a set of tools that make it easy to switch between these classes. It also allows the user to treat time series as plain data frames, facilitating the use with tools that assume rectangular data.
The package is built around a set of functions that convert time series of different classes to each other. They are frequency-agnostic, and allow the user to combine multiple non-standard and irregular frequencies. Because coercion works reliably, it is easy to write functions that work identically for all classes. So whether we want to smooth, scale, differentiate, chain-link, forecast, regularize or seasonally adjust a time series, we can use the same tsbox-command for any time series classes.
The best way to start is to check out the package website.
In the ropensci classification, this package is An improvement on other
implementations of similar algorithms in R. Many time series packages,
e.g., zoo or
tsibble contain converter
functions from one class to another. They often convert from their class
to ts objects and back, but lack converters to other time series class.
In most cases, tsbox transforms an object into an augmented data.table. And
uses the data.table infrastructure for efficient joining and reshaping. After
computation, it restores the original input class. This restoring feature is
was also used in the xts::reclass() function of the
xts package.
Christoph Sax [email protected]
Useful links:
Report bugs at https://github.com/ropensci/tsbox/issues
Copies class attributes from an existing ts-boxable series. Mainly used internally.
copy_class( x, template, preserve.mode = TRUE, preserve.names = FALSE, preserve.time = FALSE, preserve.attr = TRUE )copy_class( x, template, preserve.mode = TRUE, preserve.names = FALSE, preserve.time = FALSE, preserve.attr = TRUE )
x |
ts-boxable time series, an object of class |
template |
ts-boxable time series, an object of class |
preserve.mode |
should the mode the time column be preserved (data frame only) |
preserve.names |
should the name of the time column be preserved (data frame only) |
preserve.time |
should the values time column be preserved (data frame only) |
preserve.attr |
should the attributes of the value column be preserved (data frame only) |
Inspired by xts::reclass, which does something similar.
a ts-boxable object of the same class as template,
i.e., an object of class ts, xts, zoo,
data.frame, data.table, tbl, tbl_ts, tbl_time, tis, irts or
timeSeries.
copy_class(mdeaths, ts_tbl(fdeaths))copy_class(mdeaths, ts_tbl(fdeaths))
Mainly used internally.
relevant_class(x)relevant_class(x)
x |
ts-boxable time series, an object of class |
character, the relevant class of ts-boxable object
relevant_class(AirPassengers) x <- ts_df(AirPassengers) relevant_class(x)relevant_class(AirPassengers) x <- ts_df(AirPassengers) relevant_class(x)
ts_ turns an existing function into a function that can deal with
ts-boxable time series objects.
load_suggested(pkg) ts_(fun, class = "ts", vectorize = FALSE, reclass = TRUE) ts_apply(x, fun, ...)load_suggested(pkg) ts_(fun, class = "ts", vectorize = FALSE, reclass = TRUE) ts_apply(x, fun, ...)
pkg |
external package, to be suggested (automatically added by |
fun |
function, to be made available to all time series classes |
class |
class that the function uses as its first argument |
vectorize |
should the function be vectorized? (not yet implemented) |
reclass |
logical, should the new function return the same same ts-boxable output as imputed? |
x |
ts-boxable time series, an object of class |
... |
arguments passed to subfunction |
The ts_ function is a constructor function for tsbox time series functions.
It can be used to wrap any function that works with time series. The default
is set to R base "ts" class. ts_ deals with the conversion stuff,
'vectorizes' the function so that it can be used with multiple time series.
A function that accepts ts-boxable time series as an input.
ts_examples, for a few useful examples of functions generated by
ts_.
Vignette on how to make arbitrary functions ts-boxable.
ts_(rowSums)(ts_c(mdeaths, fdeaths)) ts_plot(mean = ts_(rowMeans)(ts_c(mdeaths, fdeaths)), mdeaths, fdeaths) ts_(function(x) predict(prcomp(x)))(ts_c(mdeaths, fdeaths)) ts_(function(x) predict(prcomp(x, scale = TRUE)))(ts_c(mdeaths, fdeaths)) ts_(dygraphs::dygraph, class = "xts") # attach series to serach path ts_attach <- ts_(attach, class = "tslist", reclass = FALSE) ts_attach(EuStockMarkets) ts_plot(DAX, SMI) detach()ts_(rowSums)(ts_c(mdeaths, fdeaths)) ts_plot(mean = ts_(rowMeans)(ts_c(mdeaths, fdeaths)), mdeaths, fdeaths) ts_(function(x) predict(prcomp(x)))(ts_c(mdeaths, fdeaths)) ts_(function(x) predict(prcomp(x, scale = TRUE)))(ts_c(mdeaths, fdeaths)) ts_(dygraphs::dygraph, class = "xts") # attach series to serach path ts_attach <- ts_(attach, class = "tslist", reclass = FALSE) ts_attach(EuStockMarkets) ts_plot(DAX, SMI) detach()
Arithmetic Operators for ts-boxable objects
e1 %ts+% e2 e1 %ts-% e2 e1 %ts*% e2 e1 %ts/% e2e1 %ts+% e2 e1 %ts-% e2 e1 %ts*% e2 e1 %ts/% e2
e1 |
ts-boxable time series, an object of class |
e2 |
ts-boxable time series, an object of class |
a ts-boxable time series, with the same class as the left input.
head(fdeaths - mdeaths) head(fdeaths %ts-% mdeaths) head(ts_df(fdeaths) %ts-% mdeaths)head(fdeaths - mdeaths) head(fdeaths %ts-% mdeaths) head(ts_df(fdeaths) %ts-% mdeaths)
Combine time series to a new, single time series. ts_bind combines time
series as they are, ts_chain chains them together, using percentage change
rates.
ts_bind(...) ts_chain(...)ts_bind(...) ts_chain(...)
... |
ts-boxable time series, an object of class |
In data frame objects, multiple time series are stored in a long data frame.
In ts and xts objects, time series are combined horizontally.
a ts-boxable object of the same class as the input, i.e., an object
of class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
If series of different classes are combined, the class of the first series
is used (if possible).
ts_c to collect multiple time series
ts_bind(ts_span(mdeaths, end = "1975-12-01"), fdeaths) ts_bind(mdeaths, c(2, 2)) ts_bind(mdeaths, 3, ts_bind(fdeaths, c(99, 2))) ts_bind(ts_dt(mdeaths), AirPassengers) # numeric vectors ts_bind(12, AirPassengers, c(2, 3)) ts_chain(ts_span(mdeaths, end = "1975-12-01"), fdeaths) ts_plot(ts_pc(ts_c( comb = ts_chain(ts_span(mdeaths, end = "1975-12-01"), fdeaths), fdeaths )))ts_bind(ts_span(mdeaths, end = "1975-12-01"), fdeaths) ts_bind(mdeaths, c(2, 2)) ts_bind(mdeaths, 3, ts_bind(fdeaths, c(99, 2))) ts_bind(ts_dt(mdeaths), AirPassengers) # numeric vectors ts_bind(12, AirPassengers, c(2, 3)) ts_chain(ts_span(mdeaths, end = "1975-12-01"), fdeaths) ts_plot(ts_pc(ts_c( comb = ts_chain(ts_span(mdeaths, end = "1975-12-01"), fdeaths), fdeaths )))
Mainly used internally.
ts_boxable(x) check_ts_boxable(x)ts_boxable(x) check_ts_boxable(x)
x |
ts-boxable time series, an object of class |
logical, either TRUE or FALSE. check_ts_boxable() fails if not
TRUE
ts_boxable(AirPassengers) ts_boxable(lm)ts_boxable(AirPassengers) ts_boxable(lm)
Collect time series as multiple time series.
ts_c(...)ts_c(...)
... |
ts-boxable time series, an object of class |
In data frame objects, multiple time series are stored in a long data frame.
In ts and xts objects, time series are combined horizontally.
a ts-boxable object of the same class as the input, i.e., an object
of class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
If series of different classes are combined, the class of the first series
is used (if possible).
ts_bind, to bind multiple time series to a single series.
ts_c(mdeaths, fdeaths) ts_c(ts_df(EuStockMarkets), AirPassengers) # labeling x1 <- ts_c( `International Airline Passengers` = ts_xts(AirPassengers), `Deaths from Lung Diseases` = ldeaths ) head(x1)ts_c(mdeaths, fdeaths) ts_c(ts_df(EuStockMarkets), AirPassengers) # labeling x1 <- ts_c( `International Airline Passengers` = ts_xts(AirPassengers), `Deaths from Lung Diseases` = ldeaths ) head(x1)
In data frame objects (data.frame, tibble, data.table), tsbox
automatically detects the time and the value column. This function changes
the column names to the defaults (time, value), so that auto-detection
can be avoided in future operations.
ts_default(x)ts_default(x)
x |
ts-boxable time series, an object of class |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
df <- ts_df(ts_c(mdeaths, fdeaths)) # non-default colnames colnames(df) <- c("id", "date", "count") # switch back to default colnames ts_default(df)df <- ts_df(ts_c(mdeaths, fdeaths)) # non-default colnames colnames(df) <- c("id", "date", "count") # switch back to default colnames ts_default(df)
In data frame objects (data.frame, tibble, data.table), tsbox
automatically detects the time and the value column. This function changes
the column names to the defaults (time, value), so that auto-detection
can be avoided in future operations.
ts_dts(x)ts_dts(x)
x |
ts-boxable time series, an object of class |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
df <- ts_df(ts_c(mdeaths, fdeaths)) # non-default colnames colnames(df) <- c("id", "date", "count") # switch back to default colnames ts_default(df)df <- ts_df(ts_c(mdeaths, fdeaths)) # non-default colnames colnames(df) <- c("id", "date", "count") # switch back to default colnames ts_default(df)
Example Functions, Generated by ts_. ts_prcomp calculates the principal
components of multiple time series, ts_dygraphs generates an interactive
graphical visualization, ts_forecast return an univariate forecast,
ts_seas the seasonally adjusted series. ts_na_interpolation imputes
missing values.
ts_prcomp(x, ...) ts_dygraphs(x, ...) ts_forecast(x, ...) ts_seas(x, ...) ts_na_interpolation(x, ...)ts_prcomp(x, ...) ts_dygraphs(x, ...) ts_forecast(x, ...) ts_seas(x, ...) ts_na_interpolation(x, ...)
x |
ts-boxable time series, an object of class |
... |
further arguments, passed to the underlying function. For help, consider these functions, e.g., stats::prcomp. |
With the exception of ts_prcomp, these functions depend on external
packages.
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
Vignette on how to make arbitrary functions ts-boxable.
ts_plot( ts_scale(ts_c( Male = mdeaths, Female = fdeaths, `First principal compenent` = -ts_prcomp(ts_c(mdeaths, fdeaths))[, 1] )), title = "Deaths from lung diseases", subtitle = "Normalized values" ) ts_plot(ts_c( male = mdeaths, female = fdeaths, ts_forecast(ts_c(`male (fct)` = mdeaths, `female (fct)` = fdeaths)) ), title = "Deaths from lung diseases", subtitle = "Exponential smoothing forecast" ) ts_plot( `Raw series` = AirPassengers, `Adjusted series` = ts_seas(AirPassengers), title = "Airline passengers", subtitle = "X-13 seasonal adjustment" ) # See ?imputeTS::na_interpolation for options dta <- ts_c(mdeaths, fdeaths) dta[c(1, 3, 10), c(1, 2)] <- NA head(ts_na_interpolation(dta, option = "spline")) ts_dygraphs(ts_c(mdeaths, EuStockMarkets))ts_plot( ts_scale(ts_c( Male = mdeaths, Female = fdeaths, `First principal compenent` = -ts_prcomp(ts_c(mdeaths, fdeaths))[, 1] )), title = "Deaths from lung diseases", subtitle = "Normalized values" ) ts_plot(ts_c( male = mdeaths, female = fdeaths, ts_forecast(ts_c(`male (fct)` = mdeaths, `female (fct)` = fdeaths)) ), title = "Deaths from lung diseases", subtitle = "Exponential smoothing forecast" ) ts_plot( `Raw series` = AirPassengers, `Adjusted series` = ts_seas(AirPassengers), title = "Airline passengers", subtitle = "X-13 seasonal adjustment" ) # See ?imputeTS::na_interpolation for options dta <- ts_c(mdeaths, fdeaths) dta[c(1, 3, 10), c(1, 2)] <- NA head(ts_na_interpolation(dta, option = "spline")) ts_dygraphs(ts_c(mdeaths, EuStockMarkets))
Replace date or time values by the first of the period. tsbox usually relies on timestamps being the first value of a period.
ts_first_of_period(x)ts_first_of_period(x)
x |
ts-boxable time series, an object of class |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
x <- ts_c( a = ts_lag(ts_df(mdeaths), "14 days"), b = ts_lag(ts_df(mdeaths), "-2 days") ) ts_first_of_period(x) ts_first_of_period(ts_lag(ts_df(austres), "14 days"))x <- ts_c( a = ts_lag(ts_df(mdeaths), "14 days"), b = ts_lag(ts_df(mdeaths), "-2 days") ) ts_first_of_period(x) ts_first_of_period(ts_lag(ts_df(austres), "14 days"))
Changes the frequency of a time series. By default, incomplete periods of regular series are omitted.
ts_frequency( x, to = c("year", "quarter", "month", "week", "day", "hour", "min", "sec"), aggregate = "mean", na.rm = FALSE )ts_frequency( x, to = c("year", "quarter", "month", "week", "day", "hour", "min", "sec"), aggregate = "mean", na.rm = FALSE )
x |
ts-boxable time series, an object of class |
to |
desired frequency, either a character string ( |
aggregate |
character string, or function. Either |
na.rm |
logical, if |
The tempdisagg package
can convert low frequency to high frequency data and
has support for ts-boxable objects. See
vignette("hf-disagg", package = "tempdisagg").
a ts-boxable time series, with the same class as the input.
ts_frequency(cbind(mdeaths, fdeaths), "year", "sum") ts_frequency(cbind(mdeaths, fdeaths), "quarter", "last") ts_frequency(AirPassengers, 4, "sum") # Note that incomplete years are omited by default ts_frequency(EuStockMarkets, "year") ts_frequency(EuStockMarkets, "year", na.rm = TRUE)ts_frequency(cbind(mdeaths, fdeaths), "year", "sum") ts_frequency(cbind(mdeaths, fdeaths), "quarter", "last") ts_frequency(AirPassengers, 4, "sum") # Note that incomplete years are omited by default ts_frequency(EuStockMarkets, "year") ts_frequency(EuStockMarkets, "year", na.rm = TRUE)
ts_ggplot() has the same syntax and produces a similar plot as ts_plot(),
but uses the ggplot2 graphic system, and
can be customized. With theme_tsbox() and scale_color_tsbox(), the output
of ts_ggplot has a similar look and feel.
ts_ggplot(..., title, subtitle, ylab = "") theme_tsbox(base_family = getOption("ts_font", ""), base_size = 12) colors_tsbox() scale_color_tsbox(...) scale_fill_tsbox(...)ts_ggplot(..., title, subtitle, ylab = "") theme_tsbox(base_family = getOption("ts_font", ""), base_size = 12) colors_tsbox() scale_color_tsbox(...) scale_fill_tsbox(...)
... |
ts-boxable time series, objects of class |
title |
title (optional) |
subtitle |
subtitle (optional) |
ylab |
ylab (optional) |
base_family |
base font family (can also be set via |
base_size |
base font size |
Both ts_plot() and ts_ggplot() combine multiple ID dimensions into a
single dimension. To plot multiple dimensions in different shapes, facets,
etc., use standard ggplot (see examples).
ts_plot(), for a simpler and faster plotting function.
ts_dygraphs(), for interactive time series plots.
# using the ggplot2 graphic system p <- ts_ggplot(total = ldeaths, female = fdeaths, male = mdeaths) p # with themes for the look and feel of ts_plot() p + theme_tsbox() + scale_color_tsbox() # also use themes with standard ggplot suppressMessages(library(ggplot2)) df <- ts_df(ts_c(total = ldeaths, female = fdeaths, male = mdeaths)) ggplot(df, aes(x = time, y = value)) + facet_wrap("id") + geom_line() + theme_tsbox() + scale_color_tsbox() ## Not run: library(dataseries) dta <- ds(c("GDP.PBRTT.A.R", "CCI.CCIIR"), "xts") ts_ggplot(ts_scale(ts_span( ts_c( `GDP Growth` = ts_pc(dta[, "GDP.PBRTT.A.R"]), `Consumer Sentiment Index` = dta[, "CCI.CCIIR"] ), start = "1995-01-01" ))) + ggplot2::ggtitle("GDP and Consumer Sentiment", subtitle = "normalized") + theme_tsbox() + scale_color_tsbox() ## End(Not run)# using the ggplot2 graphic system p <- ts_ggplot(total = ldeaths, female = fdeaths, male = mdeaths) p # with themes for the look and feel of ts_plot() p + theme_tsbox() + scale_color_tsbox() # also use themes with standard ggplot suppressMessages(library(ggplot2)) df <- ts_df(ts_c(total = ldeaths, female = fdeaths, male = mdeaths)) ggplot(df, aes(x = time, y = value)) + facet_wrap("id") + geom_line() + theme_tsbox() + scale_color_tsbox() ## Not run: library(dataseries) dta <- ds(c("GDP.PBRTT.A.R", "CCI.CCIIR"), "xts") ts_ggplot(ts_scale(ts_span( ts_c( `GDP Growth` = ts_pc(dta[, "GDP.PBRTT.A.R"]), `Consumer Sentiment Index` = dta[, "CCI.CCIIR"] ), start = "1995-01-01" ))) + ggplot2::ggtitle("GDP and Consumer Sentiment", subtitle = "normalized") + theme_tsbox() + scale_color_tsbox() ## End(Not run)
ts_index returns an indexed series, with value of 1 at the base date or
range.
ts_compound builds an index from percentage change rates, starting with 1
and compounding the rates.
ts_compound(x, denominator = 100) ts_index(x, base = NULL)ts_compound(x, denominator = 100) ts_index(x, base = NULL)
x |
ts-boxable time series, an object of class |
denominator |
positive number. Set equal to 1 if percentage change rate is given a decimal fraction |
base |
base date, character string, |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
x <- ts_pc(ts_c(fdeaths, mdeaths)) ts_compound(x) y <- ts_df(ts_c(fdeaths, mdeaths)) ts_index(y, "1974-02-01") ts_plot( `My Expert Knowledge` = ts_chain( mdeaths, ts_compound(ts_bind(ts_pc(mdeaths), 15, 23, 33)) ), `So Far` = mdeaths, title = "A Very Manual Forecast" ) # mean of 1974 = 1 ts_index(mdeaths, c("1974-01-01", "1974-12-31"))x <- ts_pc(ts_c(fdeaths, mdeaths)) ts_compound(x) y <- ts_df(ts_c(fdeaths, mdeaths)) ts_index(y, "1974-02-01") ts_plot( `My Expert Knowledge` = ts_chain( mdeaths, ts_compound(ts_bind(ts_pc(mdeaths), 15, 23, 33)) ), `So Far` = mdeaths, title = "A Very Manual Forecast" ) # mean of 1974 = 1 ts_index(mdeaths, c("1974-01-01", "1974-12-31"))
Shift time stamps in ts-boxable time series, either by a number of periods or by a fixed amount of time.
ts_lag(x, by = 1)ts_lag(x, by = 1)
x |
ts-boxable time series, an object of class |
by |
integer or character, either the number of shifting periods (integer), or an absolute amount of time (character). See details. |
The lag order, by, is defined the opposite way as in R base. Thus, -1 is a
lead and +1 a lag.
If by is integer, the time stamp is shifted by the number of periods. This
requires the series to be regular.
If by is character, the time stamp is shifted by a specific amount of time.
This can be one of one of "sec", "min", "hour", "day", "week",
"month", "quarter" or "year", optionally preceded by a (positive or
negative) integer and a space, or followed by plural "s". This is passed to
base::seq.Date(). This does not require the series to be regular.
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
ts_plot(AirPassengers, ts_lag(AirPassengers), title = "The need for glasses") ts_lag(fdeaths, "1 month") ts_lag(fdeaths, "1 year") x <- ts_df(fdeaths) ts_lag(x, "2 day") ts_lag(x, "2 min") ts_lag(x, "-1 day")ts_plot(AirPassengers, ts_lag(AirPassengers), title = "The need for glasses") ts_lag(fdeaths, "1 month") ts_lag(fdeaths, "1 year") x <- ts_df(fdeaths) ts_lag(x, "2 day") ts_lag(x, "2 min") ts_lag(x, "-1 day")
Functions to reshape multiple time series from 'wide' to 'long' and vice
versa. Note that long format data frames are ts-boxable objects, where wide
format data frames are not. ts_long automatically identifies a time
column, and uses columns on the left as id columns.
ts_long(x) ts_wide(x)ts_long(x) ts_wide(x)
x |
a ts-boxable time series, or a wide |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
x <- ts_df(ts_c(mdeaths, fdeaths)) df.wide <- ts_wide(x) df.wide ts_long(df.wide)x <- ts_df(ts_c(mdeaths, fdeaths)) df.wide <- ts_wide(x) df.wide ts_long(df.wide)
Remove NA values in ts-boxable objects, turning explicit into implicit missing values.
ts_na_omit(x)ts_na_omit(x)
x |
ts-boxable time series, an object of class |
Note that internal NAs in ts time series will not be removed, as this
conflicts with the regular structure.
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
ts_regular, for the opposite, turning implicit into explicit missing values.
x <- AirPassengers x[c(2, 4)] <- NA # A ts object does only know explicit NAs ts_na_omit(x) # by default, NAs are implicit in data frames ts_df(x) # make NAs explicit ts_regular(ts_df(x)) # and implicit again ts_na_omit(ts_regular(ts_df(x)))x <- AirPassengers x[c(2, 4)] <- NA # A ts object does only know explicit NAs ts_na_omit(x) # by default, NAs are implicit in data frames ts_df(x) # make NAs explicit ts_regular(ts_df(x)) # and implicit again ts_na_omit(ts_regular(ts_df(x)))
ts_pcy and ts_diffy calculate the percentage change rate and the
difference compared to the previous period, ts_pcy and ts_diffy calculate
the percentage change rate compared to the same period of the previous year.
ts_pca calculates annualized percentage change rates compared to the
previous period.
ts_pc(x) ts_diff(x) ts_pca(x) ts_pcy(x) ts_diffy(x)ts_pc(x) ts_diff(x) ts_pca(x) ts_pcy(x) ts_diffy(x)
x |
ts-boxable time series, an object of class |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
x <- ts_c(fdeaths, mdeaths) ts_diff(x) ts_pc(x) ts_pca(x) ts_pcy(x) ts_diffy(x)x <- ts_c(fdeaths, mdeaths) ts_diff(x) ts_pc(x) ts_pca(x) ts_pcy(x) ts_diffy(x)
Pick (and optionally rename) series from multiple time series.
ts_pick(x, ...)ts_pick(x, ...)
x |
ts-boxable time series, an object of class |
... |
character string(s), names of the series to be picked, or integer, with positions. If arguments are named, the series will be renamed. |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
# Interactive use ts_plot(ts_pick( EuStockMarkets, `My Dax` = "DAX", `My Smi` = "SMI" )) ts_pick(EuStockMarkets, c(1, 2)) ts_pick(EuStockMarkets, `My Dax` = "DAX", `My Smi` = "SMI") # Programming use to.be.picked.and.renamed <- c(`My Dax` = "DAX", `My Smi` = "SMI") ts_pick(EuStockMarkets, to.be.picked.and.renamed)# Interactive use ts_plot(ts_pick( EuStockMarkets, `My Dax` = "DAX", `My Smi` = "SMI" )) ts_pick(EuStockMarkets, c(1, 2)) ts_pick(EuStockMarkets, `My Dax` = "DAX", `My Smi` = "SMI") # Programming use to.be.picked.and.renamed <- c(`My Dax` = "DAX", `My Smi` = "SMI") ts_pick(EuStockMarkets, to.be.picked.and.renamed)
ts_plot() is a fast and simple plotting function for ts-boxable time
series, with limited customizability. For more theme options, use
ts_ggplot().
ts_plot(..., title, subtitle, ylab = "", family = getOption("ts_font", "sans"))ts_plot(..., title, subtitle, ylab = "", family = getOption("ts_font", "sans"))
... |
ts-boxable time series, an object of class |
title |
title (optional) |
subtitle |
subtitle (optional) |
ylab |
ylab (optional) |
family |
font family (optional, can also be set via |
Both ts_plot() and ts_ggplot() combine multiple ID dimensions into a
single dimension. To plot multiple dimensions in different shapes, facets,
etc., use standard ggplot.
Limited customizability of ts_plot is available via options. See examples.
ts_ggplot(), for a plotting function based on ggplot2.
ts_dygraphs(), for interactive time series plots. ts_save() to
save a plot to the file system.
ts_plot( AirPassengers, title = "Airline passengers", subtitle = "The classic Box & Jenkins airline data" ) # naming arguments ts_plot(total = ldeaths, female = fdeaths, male = mdeaths) # using different ts-boxable objects ts_plot(ts_scale(ts_c( ts_xts(airmiles), ts_tbl(co2), JohnsonJohnson, ts_df(discoveries) ))) # customize ts_plot op <- options( tsbox.lwd = 3, tsbox.col = c("gray51", "gray11"), tsbox.lty = "dashed" ) ts_plot( "Female" = fdeaths, "Male" = mdeaths ) options(op) # restore defaultsts_plot( AirPassengers, title = "Airline passengers", subtitle = "The classic Box & Jenkins airline data" ) # naming arguments ts_plot(total = ldeaths, female = fdeaths, male = mdeaths) # using different ts-boxable objects ts_plot(ts_scale(ts_c( ts_xts(airmiles), ts_tbl(co2), JohnsonJohnson, ts_df(discoveries) ))) # customize ts_plot op <- options( tsbox.lwd = 3, tsbox.col = c("gray51", "gray11"), tsbox.lty = "dashed" ) ts_plot( "Female" = fdeaths, "Male" = mdeaths ) options(op) # restore defaults
Enforces regularity in data frame and xts objects, by turning implicit
NAs into explicit NAs. In ts objects, regularity is automatically
enforced.
ts_regular(x, fill = NA)ts_regular(x, fill = NA)
x |
ts-boxable time series, an object of class |
fill |
numeric, instead of |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
x0 <- AirPassengers x0[c(10, 15)] <- NA x <- ts_na_omit(ts_dts(x0)) ts_regular(x) ts_regular(x, fill = 0) m <- mdeaths m[c(10, 69)] <- NA f <- fdeaths f[c(1, 3, 15)] <- NA ts_regular(ts_na_omit(ts_dts(ts_c(f, m))))x0 <- AirPassengers x0[c(10, 15)] <- NA x <- ts_na_omit(ts_dts(x0)) ts_regular(x) ts_regular(x, fill = 0) m <- mdeaths m[c(10, 69)] <- NA f <- fdeaths f[c(1, 3, 15)] <- NA ts_regular(ts_na_omit(ts_dts(ts_c(f, m))))
Save Previous Plot
ts_save( filename = tempfile(fileext = ".pdf"), width = 10, height = 5, device = NULL, open = TRUE )ts_save( filename = tempfile(fileext = ".pdf"), width = 10, height = 5, device = NULL, open = TRUE )
filename |
filename |
width |
width |
height |
height |
device |
device |
open |
logical, should the saved plot be opened? |
invisible TRUE, if successful
ts_plot(AirPassengers) tf <- tempfile(fileext = ".pdf") ts_save(tf) unlink(tf)ts_plot(AirPassengers) tf <- tempfile(fileext = ".pdf") ts_save(tf) unlink(tf)
Subtract mean (sum(x)/n) and divide by standard deviation
(sqrt(sum(x^2)/(n-1))). Based on base::scale().
ts_scale(x, center = TRUE, scale = TRUE)ts_scale(x, center = TRUE, scale = TRUE)
x |
ts-boxable time series, an object of class |
center |
logical |
scale |
logical |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
ts_plot(ts_scale((ts_c(airmiles, co2, JohnsonJohnson, discoveries)))) ts_plot(ts_scale(ts_c(AirPassengers, DAX = EuStockMarkets[, "DAX"])))ts_plot(ts_scale((ts_c(airmiles, co2, JohnsonJohnson, discoveries)))) ts_plot(ts_scale(ts_c(AirPassengers, DAX = EuStockMarkets[, "DAX"])))
Filter time series for a time span.
ts_span(x, start = NULL, end = NULL, template = NULL, extend = FALSE)ts_span(x, start = NULL, end = NULL, template = NULL, extend = FALSE)
x |
ts-boxable time series, an object of class |
start |
start date, character string of length 1, |
end |
end date, character string of length 1, |
template |
ts-boxable time series, an object of class |
extend |
logical. If true, the start and end values are allowed to
extend the series (by adding |
All date and times, when entered as character strings, are processed by
anytime::anydate() or anytime::anytime(). Thus a wide range of inputs are
possible. See examples.
start and end can be specified relative to each other,
using one of "sec", "min", "hour", "day", "week",
"month", "quarter" or "year", or an abbreviation. If the series are of
the same frequency, the shift can be specified in periods. See examples.
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
# use 'anytime' shortcuts ts_span(mdeaths, start = "1979") # shortcut for 1979-01-01 ts_span(mdeaths, start = "1979-4") # shortcut for 1979-04-01 ts_span(mdeaths, start = "197904") # shortcut for 1979-04-01 # it's fine to use an to date outside of series span ts_span(mdeaths, end = "2001-01-01") # use strings to set start or end relative to each other ts_span(mdeaths, start = "-7 month") # last 7 months ts_span(mdeaths, start = -7) # last 7 periods ts_span(mdeaths, start = -1) # last single value ts_span(mdeaths, end = "1e4 hours") # first 10000 hours ts_plot( ts_span(mdeaths, start = "-3 years"), title = "Three years ago", subtitle = "The last three years of available data" ) ts_ggplot( ts_span(mdeaths, end = "28 weeks"), title = "28 weeks later", subtitle = "The first 28 weeks of available data" ) + theme_tsbox() + scale_color_tsbox() # Limit span of 'discoveries' to the same span as 'AirPassengers' ts_span(discoveries, template = AirPassengers) ts_span(mdeaths, end = "19801201", extend = TRUE)# use 'anytime' shortcuts ts_span(mdeaths, start = "1979") # shortcut for 1979-01-01 ts_span(mdeaths, start = "1979-4") # shortcut for 1979-04-01 ts_span(mdeaths, start = "197904") # shortcut for 1979-04-01 # it's fine to use an to date outside of series span ts_span(mdeaths, end = "2001-01-01") # use strings to set start or end relative to each other ts_span(mdeaths, start = "-7 month") # last 7 months ts_span(mdeaths, start = -7) # last 7 periods ts_span(mdeaths, start = -1) # last single value ts_span(mdeaths, end = "1e4 hours") # first 10000 hours ts_plot( ts_span(mdeaths, start = "-3 years"), title = "Three years ago", subtitle = "The last three years of available data" ) ts_ggplot( ts_span(mdeaths, end = "28 weeks"), title = "28 weeks later", subtitle = "The first 28 weeks of available data" ) + theme_tsbox() + scale_color_tsbox() # Limit span of 'discoveries' to the same span as 'AirPassengers' ts_span(discoveries, template = AirPassengers) ts_span(mdeaths, end = "19801201", extend = TRUE)
Extract time series properties, such as the number of observations
(obs), the time differences between observations (obs), the number
of observations per year (freq), and the start time stamp (start)
and the end time stamp (end) of the series.
ts_summary(x, spark = FALSE)ts_summary(x, spark = FALSE)
x |
ts-boxable time series, an object of class |
spark |
logical should an additional column with a spark-line added to the data frame (experimental, ASCII only on Windows.) |
ts_summary returns a data.frame. Individual column can be
accessed through the $ notation (see examples).
ts_summary(ts_c(mdeaths, austres)) ts_summary(ts_c(mdeaths, austres), spark = TRUE) # Extracting specific properties ts_summary(AirPassengers)$start ts_summary(AirPassengers)$freq ts_summary(AirPassengers)$obsts_summary(ts_c(mdeaths, austres)) ts_summary(ts_c(mdeaths, austres), spark = TRUE) # Extracting specific properties ts_summary(AirPassengers)$start ts_summary(AirPassengers)$freq ts_summary(AirPassengers)$obs
Trend estimation that uses stats::loess().
ts_trend(x, ...)ts_trend(x, ...)
x |
ts-boxable time series, an object of class |
... |
arguments, passed to
|
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
Cleveland, William S., Eric Grosse, and William M. Shyu. "Local regression models." Statistical models in S. Routledge, 2017. 309-376.
ts_plot( `Raw series` = fdeaths, `Loess trend` = ts_trend(fdeaths), title = "Deaths from Lung Diseases", subtitle = "per month" )ts_plot( `Raw series` = fdeaths, `Loess trend` = ts_trend(fdeaths), title = "Deaths from Lung Diseases", subtitle = "per month" )
tsbox is built around a set of converters, which convert time series
stored as ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries to each other.
ts_data.frame(x) ts_df(x) ts_data.table(x) ts_dt(x) ts_tbl(x) ts_tibbletime(x) ts_timeSeries(x) ts_tis(x) ts_ts(x) ts_irts(x) ts_tsibble(x) ts_tslist(x) ts_xts(x) ts_zoo(x) ts_zooreg(x)ts_data.frame(x) ts_df(x) ts_data.table(x) ts_dt(x) ts_tbl(x) ts_tibbletime(x) ts_timeSeries(x) ts_tis(x) ts_ts(x) ts_irts(x) ts_tsibble(x) ts_tslist(x) ts_xts(x) ts_zoo(x) ts_zooreg(x)
x |
ts-boxable time series, an object of class |
In data frames, multiple time series will be stored in a 'long' format. tsbox detects a value, a time and zero to several id columns. Column detection is done in the following order:
Starting on the right, the first first numeric or integer column
is used as value column.
Using the remaining columns, and starting on the right again, the first
Date, POSIXct, numeric or character column is used as
time column. character strings are parsed by anytime::anytime().
The time stamp, time, indicates the beginning of a period.
All remaining columns are id columns. Each unique combination of id columns points to a time series.
Alternatively, the time column and the value column to be
explicitly named as time and value. If explicit names are used, the
column order will be ignored.
Whenever possible, tsbox relies on heuristic time conversion. When a
monthly "ts" time series, e.g., AirPassengers, is converted to a data
frame, each time stamp (of class "Date") is the first day of the month. In
most circumstances, this reflects the actual meaning of the data stored in a
"ts" object. Technically, of course, this is not correct: "ts" objects
divide time in period of equal length, while in reality, February is shorter
than January. Heuristic conversion is done for frequencies of 0.1 (decades),
1 (years), 4 (quarters) and 12 (month).
For other frequencies, e.g. 260, of EuStockMarkets, tsbox uses exact
time conversion. The year is divided into 260 equally long units, and time
stamp of a period will be a point in time (of class "POSIXct").
ts-boxable time series of the desired class, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
x.ts <- ts_c(mdeaths, fdeaths) x.ts ts_df(x.ts) suppressMessages(library(dplyr)) ts_tbl(x.ts) suppressMessages(library(data.table)) ts_dt(x.ts) suppressMessages(library(xts)) ts_xts(x.ts) # heuristic time conversion # 1 month: approx. 1/12 year ts_df(AirPassengers) # exact time conversion # 1 trading day: exactly 1/260 year ts_df(EuStockMarkets) # multiple ids a <- ts_df(ts_c(fdeaths, mdeaths)) a$type <- "level" b <- ts_pc(a) b$type <- "pc" multi.id.df <- rbind(a, b) ts_ts(multi.id.df) ts_plot(multi.id.df)x.ts <- ts_c(mdeaths, fdeaths) x.ts ts_df(x.ts) suppressMessages(library(dplyr)) ts_tbl(x.ts) suppressMessages(library(data.table)) ts_dt(x.ts) suppressMessages(library(xts)) ts_xts(x.ts) # heuristic time conversion # 1 month: approx. 1/12 year ts_df(AirPassengers) # exact time conversion # 1 trading day: exactly 1/260 year ts_df(EuStockMarkets) # multiple ids a <- ts_df(ts_c(fdeaths, mdeaths)) a$type <- "level" b <- ts_pc(a) b$type <- "pc" multi.id.df <- rbind(a, b) ts_ts(multi.id.df) ts_plot(multi.id.df)
In data frame objects (data.frame, tibble, data.table), tsbox
automatically detects the time and the value column. This function changes
the column names to the defaults (time, value), so that auto-detection
can be avoided in future operations.
ts_start(x) ts_end(x)ts_start(x) ts_end(x)
x |
ts-boxable time series, an object of class |
a ts-boxable object of the same class as x, i.e., an object of
class ts, xts, zoo, zooreg, data.frame, data.table, tbl,
tbl_ts, tbl_time, tis, irts or timeSeries.
df <- ts_df(ts_c(mdeaths, fdeaths)) # non-default colnames colnames(df) <- c("id", "date", "count") # switch back to default colnames ts_default(df)df <- ts_df(ts_c(mdeaths, fdeaths)) # non-default colnames colnames(df) <- c("id", "date", "count") # switch back to default colnames ts_default(df)