mv_impute {mvLSWimpute}

Function to apply the mvLSWimpute method and impute missing values in a multivariate locally stationary time series

Description

This function applies the mvLSWimpute method to impute missing values in a multivariate locally stationary time series. The imputation can be based on forecasts only or use information from both a forecasting and backcasting step.

Usage

mv_impute(data, p = 2, type = "forward", index = NULL)

Arguments

data

Input multivariate time series, matrix of dimension TxP where P is the number of channels and T is the length of the series.

p

The number of terms to include in the clipped predictor when carrying out one step ahead forecasting/backcasting.

type

The type of imputation to carry out, either "forward" or "forward-backward"

index

The set of time indices containing missing values, this is NULL by default and will be determined from the input series.

Value

Returns a list containing the following elements:

ImputedData

Matrix containing the imputed time series.

missing.index

Vector containing the set of time indices that have missing values.

Note

As with other time series imputation methods, mv_impute requires some data values at the start of the series. In this case, we need 5 time points.

Author(s)

Rebecca Wilson

References

Wilson, R. E., Eckley, I. A., Nunes, M. A. and Park, T. (2021) A wavelet-based approach for imputation in nonstationary multivariate time series. _Statistics and Computing_ *31* Article 18, doi:10.1007/s11222-021-09998-2.

Examples

set.seed(1)
X <- matrix(rnorm(2 * 2^8), ncol = 2)
X[1:2^7, 2] <- 3 * (X[1:2^7, 2] + 0.95 * X[1:2^7, 1])
X[-(1:2^7), 2] <- X[-(1:2^7), 2] - 0.95 * X[-(1:2^7), 1]
X[-(1:2^7), 1] <- X[-(1:2^7), 1] * 4
X <- as.ts(X)

# create some fake missing data, taking care not to have missingness hear the start of the series
missing.index = sort(sample(10:2^8, 30))

X[missing.index, ] <- NA

newdata = mv_impute(X)