Lagged {lagged}

Create Lagged objects

Description

Create objects inheriting from Lagged.

Usage

Lagged(data, ...)

Arguments

data

suitable data, see Details.

...

further arguments passed on to new().

Details

Lagged creates an object inheriting from "Lagged". The exact class depends on argument data. This is the easiest way to create lagged objects.

If data is a vector, matrix or 3D array, the result is "Lagged1d", "Lagged2d" and "Lagged3d", respectively. If data inherits from "Lagged", the result is "FlexibleLagged".

Value

a suitable "Lagged" object, as described in Details

Note

I am considering making Lagged generic.

Author(s)

Georgi N. Boshnakov

See Also

the specific classes Lagged1d, Lagged2d, Lagged3d

Examples

## a discrete distribution with outcomes 0,1,2,3,4,5,6:
v <- dbinom(0:6, size = 6, p = 0.5)
bin6 <- Lagged(v)

## names are used, if present:
names(v) <- paste0("P(x=", 0:6, ")")
bin6a <- Lagged(v)
bin6a
## subsetting drops the laggedness:
bin6a[1:3]
bin6a[]
bin6a[0:2]
## to resize (shrink or extend) the object use 'maxLag<-':
s8 <- s2 <- bin6a
maxLag(s2) <- 2
s2
## extending inserts NA's:
maxLag(s8) <- 8
s8
## use assignment to extend with specific values:
s8a <- bin6a
s8a[7:8] <- c("P(x=7)" = 0, "P(x=8)" = 0)
s8a

## adapted from examples for acf()
## acf of univariate ts
acv1 <- acf(ldeaths, plot = FALSE)
class(acv1) # "acf"
a1 <-  drop(acv1$acf) 
class(a1) # numeric

la1 <- Lagged(a1) # 1d lagged object from 'numeric':
Lagged(acv1)      # 1d lagged object from 'acf':

## acf of multivariate ts
acv2 <- acf(ts.union(mdeaths, fdeaths), plot = FALSE)
class(acv2) # "acf"
a2 <-  acv2$acf
class(a2) # 3d array
dim(a2)

la2a <- acf2Lagged(acv2) # explicitly convert 'acf' to lagged object
Lagged(acv2)            # equivalently, just use Lagged()
identical(la2a, Lagged(acv2)) # TRUE

la2a[0]    # R_0,  indexing lagged object
a2[1, , ]  # same, indexing array data from 'cf' object
acv2[0]    # same, indexing 'acf' object

la2a[1]    # R_1
a2[2, , ]  # same
acv2[1/12] # transposed, see end of this example section as to why use 1/12

la2a[2]    # R_1
a2[3, , ]  # same
acv2[2/12] # transposed, see end of this example section as to why use 1/12

## multiple lags
la2a[0:1]        # native indexing with lagged object
a2[1:2, , ]      # different ordering
acv2$acf[1:2, ,] # also different ordering

## '[' doesn't drop a dimension even when it is of length 1:
la2a[0]
la2a[1]

## to get a singleton element, use '[[':
la2a[[0]]
la2a[[1]]

## arithmetic and math operators
-la1
+la1

2*la1
la1^2
la1 + la1^2

abs(la1)
sinpi(la1)
sqrt(abs(la1))

## Summary group
max(la1)
min(la1)
range(la1)
prod(la1)
sum(la1)
any(la1 < 0)
all(la1 >= 0)

## Math2 group
round(la1)
round(la1, 2)
signif(la1)
signif(la1, 4)

## The remaining examples below are only relevant
## for users comparing to indexing of 'acf'
## 
## indexing in base R acf() is somewhat misterious, so an example with
## DIY computation of lag_1 autocovariance matrix
n <- length(mdeaths)
tmpcov <- sum((mdeaths - mean(mdeaths)) * (fdeaths - mean(fdeaths)) ) / n
msd <- sqrt(sum((mdeaths - mean(mdeaths))^2)/n)
fsd <- sqrt(sum((fdeaths - mean(fdeaths))^2)/n)
tmpcov1 <- sum((mdeaths - mean(mdeaths))[2:n] *
               (fdeaths - mean(fdeaths))[1:(n-1)] ) / n
tmpcov1 / (msd * fsd)

la2a[[1]][1,2] - tmpcov1 / (msd * fsd)  # only numerically different
## the raw acf in the 'acf' object is not surprising:
la2a[[1]][1,2] == acv2$acf[2, 1, 2] # TRUE
## ... but this probably is:
acv2[1]
## the ts here is monthly but has unit of lag 'year'
## so acv2[1] asks for lag 1 year = 12 months, thus
acv2[1/12]
all( acv2$acf[13, , ] == drop(acv2[1]$acf) )    # TRUE
all( acv2$acf[2, , ]  == drop(acv2[1/12]$acf) ) # TRUE
all( la2a[[1]]        == drop(acv2[1/12]$acf) ) # TRUE