Combines PCHIP interpolation (for gaps within observed range) with autoregressive forecasting (for future periods beyond observed data). This is the recommended entry point for complete time series imputation.
Arguments
- y
Numeric vector with potential NAs
- years
Integer vector of corresponding years (length must equal length of y)
- forecast_to
Integer: the last year to forecast to. If
NULL(default), no forecasting is performed. If provided and greater thanmax(years), the series is extended with autoregressive forecasts.
Value
List with four components:
- value
Numeric vector containing the complete imputed and/or forecasted series (extended if forecast_to > max(years))
- years
Integer vector of years corresponding to the returned values
- flag
Integer vector (0/1/2) where 0 = observed, 1 = interpolated gap, 2 = forecasted future value
- method
Character string describing which methods were applied
Examples
if (FALSE) { # \dontrun{
y <- c(10, NA, NA, 20, 22, 24)
years <- 2018:2023
# Impute internal gaps only
result1 <- impute_series(y, years)
# Impute and forecast to 2025
result2 <- impute_series(y, years, forecast_to = 2025)
} # }