towerTS function

Tower for Times Series

Fits a linear model or other regression method to to lagged elements, using the Tower approach. Here k is the number of nearest neighbors as in toweranNA. Currently predicts only the component that is just past the end of the data.

towerTS(xts,lag,k,regFtnName='lm')

Arguments

• xts: A time series.
• lag: Lag. A positive integer.
• regFtnName: Regression model to be used, currently 'lm', 'glm' (family=binomial), or 'towerKNN'.
• k: k nearest neighbors

Details

See ?towerLM and ?regtools::TStoX

Returns

Predicted value for the next item in the series.

Author(s)

Norm Matloff, Pete Mohanty

Examples

# create noisy cylcic series 
set.seed(2020)
x <- rnorm(1000)
x <- runif(1000) * sin(x) + rnorm(1000)
w <- x
# introduce 10 percent missingness
x[sample(1000, 100)] <- NA
# make predictions with lag 3 using k=4 nearest neighbors 
towerTS(x, lag=10, k=4)
#  -0.1685019