The finite geometric kernel \(w_i = \theta^i\) for lags
\(i = 0, 1, \ldots, L-1\), divided by its sum unless normalise = FALSE.
adstock_geometric() applies the same weights when max_lag is finite;
this function is exported because the weights themselves are often what you
want to plot or report.
Arguments
- max_lag
Number of periods the kernel spans, including the current period. Must be finite.
max_lag = 1means no carryover.- decay
Geometric decay coefficient in
[0, 1]. Seedecay_from_half_life().- normalise
Should the weights sum to 1? Defaults to
TRUE. See the Normalisation section ofadstock_geometric()for what this does to the interpretation of a downstream coefficient.
Examples
adstock_weights(max_lag = 5, decay = 0.5)
#> [1] 0.51612903 0.25806452 0.12903226 0.06451613 0.03225806
# Unnormalised: the current period always carries weight 1
adstock_weights(max_lag = 5, decay = 0.5, normalise = FALSE)
#> [1] 1.0000 0.5000 0.2500 0.1250 0.0625
# Normalised weights sum to 1, so the series level is preserved
sum(adstock_weights(max_lag = 10, decay = 0.7))
#> [1] 1