Manski worst-case bounds on a partially observed mean

Manskif and morie_bnd_manski document the SAME method: the mean of a partially observed outcome is bounded by filling the unseen part with the support endpoints; with treatment given, the same construction bounds each potential-outcome mean and the ATE. Rather than carry a second implementation – which would agree with the first at 1e-9 forever while doubling the surface – this function forwards to morie_bnd_manski.

Usage

Manskif(y, observed, support, treatment = NULL)

Arguments

y

Outcome; entries where observed is FALSE are ignored.

observed

Logical vector, whether the outcome was seen; ignored when treatment is given (pass NULL then).

support

Numeric length-2, the logical range of the outcome.

treatment

Optional 0/1 treatment indicator for the ATE case.

Value

Whatever morie_bnd_manski returns.

References

Manski, C. F. (1989), Anatomy of the Selection Problem, Journal of Human Resources 24(3):343-360. Manski, C. F. (1990), Nonparametric Bounds on Treatment Effects, American Economic Review Papers and Proceedings 80(2):319-323. Molinari, F. (2021), Microeconometrics with Partial Identification, Handbook of Econometrics 7A, eq. (2.11) (arXiv:2004.11751).

Examples

set.seed(7)
y <- runif(50); obs <- runif(50) > 0.2
Manskif(y, obs, support = c(0, 1))
#> $lower
#> [1] 0.4091428
#> 
#> $upper
#> [1] 0.5891428
#> 
#> $width
#> [1] 0.18
#> 
#> $p_observed
#> [1] 0.82
#> 
#> $identified
#> [1] FALSE
#> 
#> $width_identity
#> [1] "(K1 - K0)(1 - P(obs)) exactly"
#> 
#> $n
#> [1] 50
#> 
#> $method
#> [1] "Manski worst-case bounds on a partially observed mean"
#>