Computes the autocorrelation of model residuals within each trajectory,
time series or other group, for lags from 0 to max_lag. At lag 1 it is
the correlation between each residual and the residual one measurement
earlier in the same group; lag 2 skips one measurement, and so on.
Residuals are never paired across groups. Used in Advanced Statistics to
check GAMMs for autocorrelated residuals; itsadug::acf_resid() does the
same with base R graphics.
Value
A tibble with one row per lag and the columns lag and acf.
The autocorrelation at lag 0 is always 1. It is NA at lags as long as
the groups, where there are too few pairs of residuals left.
Details
The residuals must be in the order of the data, sorted by time within
each group, and res and group must have the same length: if the model
dropped rows with missing values, drop them from group too.
Examples
library(simling)
# a straight line through the PRICE trajectories: the residuals of
# neighbouring measurements are strongly correlated
m <- lm(f2 ~ time, data = price_f2)
resid_acf(residuals(m), price_f2$traj)
#> # A tibble: 9 × 2
#> lag acf
#> <int> <dbl>
#> 1 0 1
#> 2 1 0.887
#> 3 2 0.647
#> 4 3 0.370
#> 5 4 0.164
#> 6 5 0.108
#> 7 6 0.178
#> 8 7 0.269
#> 9 8 0.307
