Skip to contents

Extract results from a sensitivity analysis

Source: R/accessors.R
regsensitivity-accessors.Rd

Accessors for the quantities most often reported, so that scripts do not depend on how the result objects are laid out internally.

Usage

breakdown_point(x, ...)

# S3 method for class 'regsensitivity'
breakdown_point(x, signed = FALSE, ...)

# S3 method for class 'regsensitivity_boot'
breakdown_point(x, signed = FALSE, ...)

# S3 method for class 'regsensitivity_multi'
breakdown_point(x, ...)

breakdown_frontier(x, ...)

# S3 method for class 'regsensitivity'
breakdown_frontier(x, signed = FALSE, ...)

identified_set(x, ...)

# S3 method for class 'regsensitivity'
identified_set(x, ...)

sensitivity_params(x, ...)

# S3 method for class 'regsensitivity'
sensitivity_params(x, ...)

# S3 method for class 'regsensitivity'
coef(object, ...)

# S3 method for class 'regsensitivity'
nobs(object, ...)

Arguments

x, object

A result object.

...

Ignored.

signed

Logical. The breakdown point is reported as a magnitude, which is the number quoted as "the breakdown point". With signed = TRUE an Oster breakdown keeps the sign that records the direction of selection; see regsen_breakdown().

Value

See Details.

Details

  • breakdown_point() returns the breakdown point. For regsen_bounds() output it is a scalar (NA when the sweep does not define one); for regsen_breakdown() output it is one value per point of the frontier, in the order of breakdown_frontier(); for regsen_boot() output, the point estimate; for regsen_multi() output, one value per treatment, named by treatment.

  • breakdown_frontier() returns the frontier as a data frame whose first column is named after the parameter swept (cbar, beta, rxbar, r2long or maxovb) and whose second is breakdown.

  • identified_set() returns the bounds of a regsen_bounds() sweep: one row per point, with the sensitivity parameters and bmin, bmax.

  • sensitivity_params() returns the sensitivity parameters the analysis was run at, as a named list.

  • coef() returns the short- and medium-regression coefficients on the treatment, the two estimates the sensitivity analysis starts from, and nobs() the number of observations used.

See also

as.data.frame.regsensitivity() for the full results table.

Examples

data(bfg2020)
f <- avgrep2000to2016 ~ tye_tfe890_500kNI_100_l6 + log_area_2010 + lat + lon
w1 <- c("log_area_2010", "lat", "lon")

bnds <- regsen_bounds(f, bfg2020, compare = w1, cbar = 0.1)
breakdown_point(bnds)
#> [1] 0.3905962
head(identified_set(bnds))
#>       rxbar rybar cbar       bmin     bmax
#> 1 0.0000000   Inf  0.1  1.3854162 1.385416
#> 2 0.2377361   Inf  0.1  0.5599466 2.210886
#> 3 0.4754721   Inf  0.1 -0.3235995 3.094432
#> 4 0.7132082   Inf  0.1 -1.2922838 4.063116
#> 5 0.9509442   Inf  0.1 -2.3863884 5.157221
#> 6 1.1886803   Inf  0.1 -3.6708919 6.441724
coef(bnds)
#>    short   medium 
#> 1.707838 1.385416 

bk <- regsen_breakdown(f, bfg2020, compare = w1, cbar = c(0, 0.5, 1))
breakdown_frontier(bk)
#>   cbar breakdown
#> 1  0.0 0.4044354
#> 2  0.5 0.3749327
#> 3  1.0 0.3749327
breakdown_point(bk)
#> [1] 0.4044354 0.3749327 0.3749327