Apply any R function on rolling windows
-Dawid Kałędkowski
+Dawid +Kałędkowski
-2022-09-13
+2024-03-04
Source:vignettes/apply_any_r_function.Rmd
apply_any_r_function.Rmd2022-09-13
Using runner
-runner package provides functions applied on running windows. The most universal function is runner::runner which gives user possibility to apply any R function f on running windows. Running windows are defined for each data window size k, lag with respect to their indexes. Unlike other available R packages, runner supports any input and output type and also gives full control to manipulate window size and lag/lead.
There are different kinds of running windows and all of them are implemented in runner.
runner package provides functions applied on running
+windows. The most universal function is runner::runner
+which gives user possibility to apply any R function f on
+running windows. Running windows are defined for each data window size
+k, lag with respect to their indexes. Unlike
+other available R packages, runner supports any input and
+output type and also gives full control to manipulate window size and
+lag/lead.
There are different kinds of running windows and all of them are
+implemented in runner.
Cumulative windows
-The simplest window type which is similar to base::cumsum. At each element window is defined by all elements appearing before current.
The simplest window type which is similar to
+base::cumsum. At each element window is defined by all
+elements appearing before current.
In runner this can be achieved as simple by:
@@ -134,10 +145,15 @@Cumulative windows
Constant sliding windows
-Second type of windows are these commonly known as running/rolling/moving/sliding windows. This types of windows moves along the index instead of cumulating like a previous one.
+
-Following diagram illustrates running windows of lengthk = 4. Each of 15 windows contains 4 elements (except first three).Second type of windows are these commonly known as +running/rolling/moving/sliding windows. This types of windows moves +along the index instead of cumulating like a previous one.
+Following diagram illustrates running windows of length +k = 4. Each of 15 windows contains 4 elements (except first +three).-
To obtain constant sliding windows one just needs to specify
+kargumentTo obtain constant sliding windows one just needs to specify +
kargument@@ -100,15 +100,32 @@# summarizing - sum of 4-elements runner( @@ -164,7 +180,18 @@Constant sliding windows
Windows depending on date
-By default
runnercalculates on assumption that index increments by one, but sometimes data points in dataset are not equally spaced (missing weekends, holidays, other missings) and thus window size should vary to keep expected time frame. If one specifiesidxargument, than running functions are applied on windows depending on date rather on a sequence 1-n.idxshould be the same length asxand should be of typeDate,POSIXtorinteger. Example below illustrates window of sizek = 5lagged bylag = 1. Note that one can specify alsok = "5 days"andlag = "day"as inseq.POSIXt.
+By default
runnercalculates on assumption that index +increments by one, but sometimes data points in dataset are not equally +spaced (missing weekends, holidays, other missings) and thus window size +should vary to keep expected time frame. If one specifies +idxargument, than running functions are applied on windows +depending on date rather on a sequence 1-n.idxshould be +the same length asxand should be of type +Date,POSIXtorinteger. Example +below illustrates window of sizek = 5lagged by +lag = 1. Note that one can specify also +k = "5 days"andlag = "day"as in +seq.POSIXt.
In the example below in square brackets ranges for each window.
diff --git a/articles/built-in_functions.html b/articles/built-in_functions.html index 5cc4dc9..8eb0601 100644 --- a/articles/built-in_functions.html +++ b/articles/built-in_functions.html @@ -39,7 +39,7 @@@@ -172,7 +199,7 @@Windows depending on date # summarize - mean runner::runner( - x = idx, + x = idx, k = 5, # 5-days window lag = 1, idx = idx, @@ -182,7 +209,7 @@
Windows depending on date # use Date or datetime sequences runner::runner( - x = idx, + x = idx, k = "5 days", # 5-days window lag = 1, idx = Sys.Date() + idx, @@ -191,7 +218,7 @@
Windows depending on date # obtain window from above illustration runner::runner( - x = idx, + x = idx, k = "5 days", lag = 1, idx = Sys.Date() + idx @@ -200,14 +227,20 @@
Windows depending on date
running at
-Runner by default returns vector of the same size as
+xunless one puts any-size vector toatargument. Each element ofatis an index on which runner calculates function. Example below illustrates output of runner forat = c(13, 27, 45, 31)which gives windows in ranges enclosed in square brackets. Range forat = 27is[22, 26]which is not available in current indices.Runner by default returns vector of the same size as
x+unless one puts any-size vector toatargument. Each +element ofatis an index on which runner calculates +function. Example below illustrates output of runner for +at = c(13, 27, 45, 31)which gives windows in ranges +enclosed in square brackets. Range forat = 27is +[22, 26]which is not available in current indices.
-idx <- c(4, 6, 7, 13, 17, 18, 18, 21, 27, 31, 37, 42, 44, 47, 48) # summary runner::runner( - x = 1:15, + x = 1:15, k = 5, lag = 1, idx = idx, @@ -217,13 +250,21 @@running at# full window runner::runner( - x = idx, + x = idx, k = 5, lag = 1, idx = idx, at = c(18, 27, 48, 31) )
+
atcan also be specified as interval of the output defined by time interval which results in obtaining results on following indicesseq(min(idx), max(idx), by = "<time interval>"). Interval can be set in the same way as inseq.POSIXtfunction. It’s worth noting thatatinterval shouldn’t be more frequent than interval ofidx- forDatethe most frequent interval is a"day", forPOSIXtit’s a"sec".
atcan also be specified as interval of the output +defined by time interval which results in obtaining results on following +indices +seq(min(idx), max(idx), by = "<time interval>"). +Interval can be set in the same way as inseq.POSIXt+function. It’s worth noting thatatinterval shouldn’t be +more frequent than interval ofidx- forDate+the most frequent interval is a"day", for +POSIXtit’s a"sec".@@ -420,7 +508,7 @@idx_date <- seq(Sys.Date(), Sys.Date() + 365, by = "1 month") @@ -251,20 +292,30 @@running at
Move and stretch window in time
-One can stretch window length by
+kand shift in time (or index) usinglag. Both arguments can beintegerand also time interval like for example2 months. Ifkorlagare a single value then window size/lag are constant for all elements of x. User can also specifyk/lagas vector, then size and lag will vary for each window. Bothkandlagcan be oflength(.) == 1,length(.) == length(x)orlength(.) == length(at)(ifatis specified).lagcan be negative and positive whilekonly non-negative.One can stretch window length by
kand shift in time (or +index) usinglag. Both arguments can be +integerand also time interval like for example +2 months. Ifkorlagare a +single value then window size/lag are constant for all elements of x. +User can also specifyk/lagas vector, then size and lag +will vary for each window. Bothkandlagcan +be oflength(.) == 1,length(.) == length(x)+orlength(.) == length(at)(ifatis +specified).lagcan be negative and positive while +konly non-negative.@@ -288,16 +341,27 @@# summarizing - concatenating runner::runner( - x = 1:10, - lag = c(-1, 2, -1, -2, 0, 0, 5, -5, -2, -3), - k = c(0, 1, 1, 1, 1, 5, 5, 5, 5, 5), + x = 1:10, + lag = c(-1, 2, -1, -2, 0, 0, 5, -5, -2, -3), + k = c(0, 1, 1, 1, 1, 5, 5, 5, 5, 5), f = paste, collapse = "," ) # full window runner::runner( - x = 1:10, + x = 1:10, lag = 1, k = c(1, 1, 1, 1, 1, 5, 5, 5, 5, 5) ) @@ -273,13 +324,15 @@Move and stretch window in timeidx <- c(4, 6, 7, 13, 17, 18, 18, 21, 27, 31, 37, 42, 44, 47, 48) runner::runner( - x = 1:15, - lag = sample(c("-2 days", "-1 days", "1 days", "2 days"), - size = 15, - replace = TRUE), - k = sample(c("5 days", "10 days", "15 days"), - size = 15, - replace = TRUE), + x = 1:15, + lag = sample(c("-2 days", "-1 days", "1 days", "2 days"), + size = 15, + replace = TRUE + ), + k = sample(c("5 days", "10 days", "15 days"), + size = 15, + replace = TRUE + ), idx = Sys.Date() + idx, f = function(x) mean(x) )
Move and stretch window in time
-NApaddingUsing
+runnerone can also specifyna_pad = TRUEwhich would returnNAfor any window which is partially out of range - meaning that there is no sufficient number of observations to fill the window. By defaultna_pad = FALSE, which means that incomplete windows are calculated anyway.na_padis applied on normal cumulative windows and on windows depending on date. In example below two windows exceed range given byidxso for these windows are empty forna_pad = TRUE. If used setsna_pad = FALSEfirst window will be empty (no single element within[-2, 3]) and last window will return elements within matchingidx.Using
runnerone can also specify +na_pad = TRUEwhich would returnNAfor any +window which is partially out of range - meaning that there is no +sufficient number of observations to fill the window. By default +na_pad = FALSE, which means that incomplete windows are +calculated anyway.na_padis applied on normal cumulative +windows and on windows depending on date. In example below two windows +exceed range given byidxso for these windows are empty +forna_pad = TRUE. If used setsna_pad = FALSE+first window will be empty (no single element within +[-2, 3]) and last window will return elements within +matchingidx.
idx <- c(4, 6, 7, 13, 17, 18, 18, 21, 27, 31, 37, 42, 44, 47, 48) runner::runner( - x = 1:15, - k = 5, - lag = 1, - idx = idx, + x = 1:15, + k = 5, + lag = 1, + idx = idx, at = c(4, 18, 48, 51), na_pad = TRUE, f = function(x) mean(x) @@ -306,12 +370,17 @@
Using runner with
-data.frameUser can also put
+data.frameintoxargument and apply functions which involve multiple columns. In example below we calculate beta parameter oflmmodel on 1, 2, …, n observations respectively. On the plot one can observe howlmparameter adapt with increasing number of observation.User can also put
data.frameintox+argument and apply functions which involve multiple columns. In example +below we calculate beta parameter oflmmodel on 1, 2, …, n +observations respectively. On the plot one can observe how +lmparameter adapt with increasing number of +observation.-x <- cumsum(rnorm(40)) y <- 3 * x + rnorm(40) date <- Sys.Date() + cumsum(sample(1:3, 40, replace = TRUE)) # unequaly spaced time series -group <- rep(c("a", "b"), 20) +group <- rep(c("a", "b"), 20) df <- data.frame(date, group, y, x) @@ -323,7 +392,12 @@Using runner with
data.frame< ) plot(slope)One can also use
+runnerwithdplyralso with problematicgroup_byoperations, without need to apply group_modify. Below we apply grouped 20-days beta, by specifying window lengthk = "10 days"and providing column name where indices (dates) are kept.One can also use
runnerwithdplyralso +with problematicgroup_byoperations, without need to apply +group_modify. +Below we apply grouped 20-days beta, by specifying window length +k = "10 days"and providing column name where indices +(dates) are kept.-library(dplyr) @@ -344,7 +418,11 @@Using runner with
data.frame< summ %>% ggplot(aes(x = date, y = cumulative_mse, group = group, color = group)) + geom_line()When user executes multiple
+runnercalls indplyrmutate, one can also userun_byfunction to prespecify arguments intidyversepipeline. In the example belowrunnerfunctions are applied onk = "20 days"calculated on"date"column.When user executes multiple
runnercalls in +dplyrmutate, one can also userun_byfunction +to prespecify arguments intidyversepipeline. In the +example belowrunnerfunctions are applied on +k = "20 days"calculated on"date"column.df %>% group_by(group) %>% @@ -353,29 +431,30 @@Using runner with
data.frame< cumulative_mse = runner( x = ., f = function(x) { - mean((residuals(lm(y ~ x, data = x))) ^ 2) + mean((residuals(lm(y ~ x, data = x)))^2) } ), - intercept = runner( x = ., f = function(x) { coefficients(lm(y ~ x, data = x))[1] } ), - slope = runner( x = ., f = function(x) { coefficients(lm(y ~ x, data = x))[2] } - ) + ) )Parallel mode
-The
+runnerfunction can also compute windows in parallel mode. The function doesn’t initialize the parallel cluster automatically but one have to do this outside and pass it to therunnerthroughclargument.The
runnerfunction can also compute windows in parallel +mode. The function doesn’t initialize the parallel cluster automatically +but one have to do this outside and pass it to therunner+throughclargument.-library(parallel) @@ -391,14 +470,23 @@Parallel mode) stopCluster(cl)
Executing
+runnerin parallel mode isn’t always faster than a single thread. Multiple-thread computation generates some overhead due to managing the nodes. In general, complex functions which bases on processor (e.g. loops) used to be quicker in parallel mode but one should assess itself which option has the edge in specific situation.Executing
runnerin parallel mode isn’t always +faster than a single thread. Multiple-thread computation +generates some overhead due to managing the nodes. In general, +complex functions which bases on processor (e.g. loops) used to be +quicker in parallel mode but one should assess itself which option +has the edge in specific situation.Build-in functions
-With
+runnerone can use any R functions, but some of them are optimized for speed reasons. These functions are:
-- aggregating functions -length_run,min_run,max_run,minmax_run,sum_run,mean_run,streak_run
-- utility functions -fill_run,lag_run,which_runWith
runnerone can use any R functions, but some of +them are optimized for speed reasons. These functions are:
+- aggregating functions -length_run,min_run, +max_run,minmax_run,sum_run, +mean_run,streak_run
+- utility functions -fill_run,lag_run, +which_runBuild-in functions -
Site built with pkgdown 2.0.6.
+Site built with pkgdown 2.0.7.
Built-in functions
-This tutorial presents built-in functions in runner package which goal is to maximize performance. Even if one can apply any R function with
+runner::runner, built-in functions are multiple times faster than R equivalent. Before you proceed further to this tutorial, make sure you know what the “running functions are”.This tutorial presents built-in functions in runner package which +goal is to maximize performance. Even if one can apply any R function +with
runner::runner, built-in functions are multiple times +faster than R equivalent. Before you proceed further to this tutorial, +make sure you know what the “running functions +are”.Running aggregations
Running
-<mean,sum,min,max>_runRunner provides basic aggregation methods calculated within running windows. Below example showing some functions behavior for different arguments setup.
+min_runcalculates current minimum for all elements of the vector. Let’s take a look at 8’th element of the vector whichmin_runis calculated on.
-First setup uses default values, so algorithm is looking for minimum value in all elements before actual (i=8). By default missing values are removed before calculations by argumentna_rm = TRUE, and also window is not specified. The default is equivalent ofbase::cumminwith additional option to ignoreNAvalues. In second example within window k=5, the lowest value is -3. In the last example minimum is not available due to existence ofNA. Graphical example is reproduced below in the code.Runner provides basic aggregation methods calculated within running +windows. Below example showing some functions behavior for different +arguments setup.
min_runcalculates current minimum for all +elements of the vector. Let’s take a look at 8’th element of the vector +whichmin_runis calculated on.
+First setup uses default values, so algorithm is looking for minimum +value in all elements before actual (i=8). By default missing values are +removed before calculations by argumentna_rm = TRUE, and +also window is not specified. The default is equivalent of +base::cumminwith additional option to ignore +NAvalues. In second example within window k=5, the lowest +value is -3. In the last example minimum is not available due to +existence ofNA. Graphical example is reproduced below in +the code.
+ narm_f = min_run(x, na_rm = FALSE) +)library(runner) @@ -116,9 +133,10 @@Running
<mean,sum,min,max>_run data.frame( x, - default = min_run(x, na_rm = TRUE), + default = min_run(x, na_rm = TRUE), k_5 = min_run(x, k = 5, na_rm = TRUE), - narm_f = min_run(x, na_rm = FALSE))-## x default k_5 narm_f ## 1 1 1 1 1 ## 2 -5 -5 -5 -5 @@ -132,21 +150,33 @@Running
<mean,sum,min,max>_run## 10 NA -5 -1 NA ## 11 -2 -5 -2 NA ## 12 3 -5 -2 NAIn above example constant
+k = 5has been used which means that for each element, current minimum is calculated on last 5-elements. It may happen that one can have time series where elements are not equally spaced in time, which effects ink = 5not constant. In example below 5-days sum is calculated. To achieve this, one should put date variable toidxargument.
-Illustration below shows two sums calculated in 5-days window span. In both cases 5-days fit in 3-elements windows. Equivalent R code below.In above example constant
k = 5has been used which +means that for each element, current minimum is calculated on last +5-elements. It may happen that one can have time series where elements +are not equally spaced in time, which effects ink = 5not +constant. In example below 5-days sum is calculated. To achieve this, +one should put date variable toidxargument.
+Illustration below shows two sums calculated in 5-days window span. In +both cases 5-days fit in 3-elements windows. Equivalent R code +below.
x <- c(-0.5910, 0.0266, -1.5166, -1.3627, 1.1785, -0.9342, 1.3236, 0.6249) -idx <- as.Date(c("1970-01-03", "1970-01-06", "1970-01-09", "1970-01-12", - "1970-01-13", "1970-01-16", "1970-01-17", "1970-01-19")) +idx <- as.Date(c( + "1970-01-03", "1970-01-06", "1970-01-09", "1970-01-12", + "1970-01-13", "1970-01-16", "1970-01-17", "1970-01-19" +)) sum_run(x, k = 5, idx = idx)-## [1] -0.5910 -0.5644 -1.4900 -2.8793 -1.7008 -1.1184 1.5679 1.0143Specifying
+lagargument shift of the window by number of elements or time periods (ifidxis specified).Specifying
lagargument shift of the window by number of +elements or time periods (ifidxis specified).x <- c(-0.5910, 0.0266, -1.5166, -1.3627, 1.1785, -0.9342, 1.3236, 0.6249) -idx <- as.Date(c("1970-01-03", "1970-01-06", "1970-01-09", "1970-01-12", - "1970-01-13", "1970-01-16", "1970-01-17", "1970-01-19")) +idx <- as.Date(c( + "1970-01-03", "1970-01-06", "1970-01-09", "1970-01-12", + "1970-01-13", "1970-01-16", "1970-01-17", "1970-01-19" +)) sum_run(x, k = 5, lag = 2, idx = idx)@@ -154,15 +184,23 @@## [1] NA -0.5910 -0.5644 -1.4900 -1.5166 -0.1842 -0.1842 1.5679Running
<mean,sum,min,max>_runRunning streak
-To count consecutive elements in specified window one can use
+streak_run. Following figure illustrates how streak is calculated with three different options setup for 9th element of the input vectorx. First shows default configuration, with full window andna_rm = TRUE. Second example count withink = 4window with count reset onNA. Last example counting streak with continuation afterNA. Visualization also supported with corresponding R code.To count consecutive elements in specified window one can use +
streak_run. Following figure illustrates how streak is +calculated with three different options setup for 9th element of the +input vectorx. First shows default configuration, with +full window andna_rm = TRUE. Second example count within +k = 4window with count reset onNA. Last +example counting streak with continuation afterNA. +Visualization also supported with corresponding R code.
+ s2 = streak_run(x, k = 4) +)x <- c("A", "B", "A", "A", "B", "B", "B", NA, "B", "B", "A", "B") data.frame( - x, + x, s0 = streak_run(x), s1 = streak_run(x, k = 4, na_rm = FALSE), - s2 = streak_run(x, k = 4))-## x s0 s1 s2 ## 1 A 1 1 1 ## 2 B 1 1 1 @@ -176,17 +214,25 @@Running streak## 10 B 5 NA 3 ## 11 A 1 1 1 ## 12 B 1 1 1
Streak is often used in sports to count number of wins or loses of the team/player. To count consecutive wins or loses in 5-days period, one have to specify
+k = 5and include dates intoidxargument. Specifyinglagshifts window bounds by number of elements or time periods (ifidxis specified).Streak is often used in sports to count number of wins or loses of +the team/player. To count consecutive wins or loses in 5-days period, +one have to specify
k = 5and include dates into +idxargument. Specifyinglagshifts window +bounds by number of elements or time periods (ifidxis +specified).+ streak_5d_lag = streak_run(x, k = 5, lag = 1, idx = idx) +)x <- c("W", "W", "L", "L", "L", "W", "L", "L") -idx <- as.Date(c("2019-01-03", "2019-01-06", "2019-01-09", "2019-01-12", - "2019-01-13", "2019-01-16", "2019-01-17", "2019-01-19")) +idx <- as.Date(c( + "2019-01-03", "2019-01-06", "2019-01-09", "2019-01-12", + "2019-01-13", "2019-01-16", "2019-01-17", "2019-01-19" +)) data.frame( idx, x, streak_5d = streak_run(x, k = 5, idx = idx), - streak_5d_lag = streak_run(x, k = 5, lag = 1, idx = idx))
## idx x streak_5d streak_5d_lag
## 1 2019-01-03 W 1 NA
## 2 2019-01-06 W 2 1
@@ -204,11 +250,18 @@ Utility functions
Improved lag
-Idea of lag_run is the same as well known stats::lag, with distinction that lag_run can depend on time or any other indexes passed to idx argument. This means that lag_run can shift by lag elements of the vector or by lag time periods (if idx is specified).
+Idea of lag_run is the same as well known
+stats::lag, with distinction that lag_run can
+depend on time or any other indexes passed to idx argument.
+This means that lag_run can shift by lag
+elements of the vector or by lag time periods (if
+idx is specified).
x <- c(-0.5910, 0.0266, -1.5166, -1.3627, 1.1785, -0.9342, 1.3236, 0.6249)
-idx <- as.Date(c("1970-01-03", "1970-01-06", "1970-01-09", "1970-01-12",
- "1970-01-13", "1970-01-16", "1970-01-17", "1970-01-19"))
+idx <- as.Date(c(
+ "1970-01-03", "1970-01-06", "1970-01-09", "1970-01-12",
+ "1970-01-13", "1970-01-16", "1970-01-17", "1970-01-19"
+))
lag_run(x, lag = 3, idx = idx)
## [1] NA -0.5910 0.0266 -1.5166 NA 1.1785 NA -0.9342
@@ -216,14 +269,28 @@ Improved lag
Filling missing values
-Function used to replace NA with previous non-NA element. To understand how fill_run works, take a look on illustration. Row ‘x’ represents input, and another rows represent output with NA replaced by fill_run with different options setup (run_for_first = TRUE and only_within = TRUE respectively). By default, fill_run replaces all NA if they were preceded by any value. If NA appeared in the beginning of the vector then it would not be replaced. But if user specify run_for_first = TRUE initial empty values values will be replaced by next non-empty value. Option only_within = TRUE means that NA values would be replaced if they were surrounded by pair of identical values. No windows provided in this functionality.
+Function used to replace NA with previous non-NA
+element. To understand how fill_run works, take a look on
+illustration. Row ‘x’ represents input, and another rows represent
+output with NA replaced by fill_run with
+different options setup (run_for_first = TRUE and
+only_within = TRUE respectively). By default,
+fill_run replaces all NA if they were preceded
+by any value. If NA appeared in the beginning of the vector
+then it would not be replaced. But if user specify
+run_for_first = TRUE initial empty values values will be
+replaced by next non-empty value. Option only_within = TRUE
+means that NA values would be replaced if they were
+surrounded by pair of identical values. No windows provided in this
+functionality.
x <- c(NA, NA, "b", "b", "a", NA, NA, "a", "b", NA, "a", "b")
-data.frame(x,
- f1 = fill_run(x),
- f2 = fill_run(x,run_for_first = TRUE),
- f3 = fill_run(x, only_within = TRUE))
+data.frame(x,
+ f1 = fill_run(x),
+ f2 = fill_run(x, run_for_first = TRUE),
+ f3 = fill_run(x, only_within = TRUE)
+)
## x f1 f2 f3
## 1 <NA> <NA> b <NA>
## 2 <NA> <NA> b <NA>
@@ -241,15 +308,25 @@ Filling missing values
Running which
-To obtain index number of element matching some condition in window, one can use which_run, which returns index of TRUE element appeared before n-th element of a vector. If na_rm = TRUE is specified, missing is treated as FALSE, and is ignored while searching for TRUE. While user set na_rm = FALSE like in second example, function returns NA, because in following window TRUE appears after missing and it’s impossible to be certain which is first (missing is an element of unknown value - could be TRUE or FALSE).
+To obtain index number of element matching some condition in window,
+one can use which_run, which returns index of
+TRUE element appeared before n-th element of a vector. If
+na_rm = TRUE is specified, missing is treated as
+FALSE, and is ignored while searching for
+TRUE. While user set na_rm = FALSE like in
+second example, function returns NA, because in following
+window TRUE appears after missing and it’s impossible to be
+certain which is first (missing is an element of unknown value - could
+be TRUE or FALSE).
x <- c(T, T, T, F, NA, T, F, NA, T, F, T, F)
data.frame(
- x,
+ x,
s0 = which_run(x, which = "first"),
s1 = which_run(x, na_rm = FALSE, k = 5, which = "first"),
- s2 = which_run(x, k = 5, which = "last"))
+ s2 = which_run(x, k = 5, which = "last")
+)
## x s0 s1 s2
## 1 TRUE 1 1 1
## 2 TRUE 1 1 2
@@ -263,7 +340,12 @@ Running which## 10 FALSE 1 6 9
## 11 TRUE 1 NA 11
## 12 FALSE 1 NA 11
which argument (‘first’ or ‘last’) used with which_run determines which index of matching element should be returned from window. In below illustration in k = 4 elements window there are two TRUE values, and depending on which argument output is equal 2 or 4.
which argument (‘first’ or ‘last’) used with
+which_run determines which index of matching element should
+be returned from window. In below illustration in k = 4
+elements window there are two TRUE values, and depending on
+which argument output is equal 2 or
+4.
