roll_conv: Calculate the rolling convolutions (weighted sums) of a _time...

In algoquant/HighFreq: High Frequency Time Series Management

Calculate the rolling convolutions (weighted sums) of a time series with a single-column matrix of weights.

Description

Calculate the rolling convolutions (weighted sums) of a time series with a single-column matrix of weights.

Usage

roll_conv(tseries, weightv)

Arguments

tseries	A time series or a matrix of data.
weightv	A single-column matrix of weights.

Details

The function roll_conv() calculates the convolutions of the matrix columns with a single-column matrix of weights. It performs a loop over the matrix rows and multiplies the past (higher) values by the weights. It calculates the rolling weighted sums of the past data.

The function roll_conv() uses the Armadillo function arma::conv2(). It performs a similar calculation to the standard R function
filter(x=tseries, filter=weightv, method="convolution", sides=1), but it's over 6 times faster, and it doesn't produce any leading NA values.

Value

A matrix with the same dimensions as the input argument tseries.

Examples

## Not run: 
# First example
# Calculate a time series of returns
retp <- na.omit(rutils::etfenv$returns[, c("IEF", "VTI")])
# Create simple weights equal to a 1 value plus zeros
weightv <- c(1, rep(0, 10))
# Calculate rolling weighted sums
retf <- HighFreq::roll_conv(retp, weightv)
# Compare with original
all.equal(coredata(retp), retf, check.attributes=FALSE)
# Second example
# Calculate exponentially decaying weights
weightv <- exp(-0.2*(1:11))
weightv <- weightv/sum(weightv)
# Calculate rolling weighted sums
retf <- HighFreq::roll_conv(retp, weightv)
# Calculate rolling weighted sums using filter()
retc <- filter(x=retp, filter=weightv, method="convolution", sides=1)
# Compare both methods
all.equal(retc[-(1:11), ], retf[-(1:11), ], check.attributes=FALSE)

## End(Not run)

algoquant/HighFreq documentation built on Feb. 9, 2024, 8:15 p.m.