Add fitted means to the covariate table

Fitted means belong to covariate occasions. The estimating equation evaluates the link at each observed covariate vector, and the kernel weight is what ties that vector to a response occasion. So augment() returns data_x with a .fitted column, \(g(X_i(S_{ik})^\top \hat\beta)\).

Usage

# S3 method for class 'kee_async'
augment(x, data_x, ...)

Arguments

x

A kee_async fit.

data_x

The covariate table the fit was made from.

...

Unused.

Value

data_x as a tibble with a .fitted column added.

Details

There is deliberately no .resid. A residual needs a response value at \(S_{ik}\), and asynchronous data has none; any residual reported here would have to be invented.

The fit does not retain its data, so data_x must be supplied.

Examples

set.seed(1)
d <- sim_async_data(n = 150)
fit <- kee_async(d$y, d$x, y ~ x, id = id, time = time, h = 0.3)
augment(fit, d$x)
#> # A tibble: 742 × 4
#>       id   time       x .fitted
#>    <int>  <dbl>   <dbl>   <dbl>
#>  1     1 0.0618 -0.864   -0.595
#>  2     1 0.629   0.345    1.12 
#>  3     1 0.661   0.437    1.25 
#>  4     1 0.945   0.253    0.985
#>  5     2 0.0233  0.292    1.04 
#>  6     2 0.477   0.716    1.64 
#>  7     2 0.530   0.827    1.80 
#>  8     2 0.553   0.974    2.00 
#>  9     2 0.732   0.0543   0.704
#> 10     2 0.789  -0.101    0.484
#> # ℹ 732 more rows