PPI++ Paired t-Test

Performs a paired hypothesis test for the mean difference using PPI++. This is equivalent to a one-sample test on the differences D = Y_A - Y_B.

Usage

ppi_paired_test(D_L, fD_L, fD_U, delta0 = 0, alpha = 0.05, lambda = NULL)

Arguments

D_L

Numeric vector of labeled differences (Y_A - Y_B).

fD_L

Numeric vector of predicted differences for labeled samples.

fD_U

Numeric vector of predicted differences for unlabeled samples.

delta0

Null hypothesis value for the mean difference (default 0).

alpha

Significance level (default 0.05).

lambda

Optional lambda. If NULL, uses EIF-optimal.

Value

A list with components (same as ppi_mean_test).

Examples

set.seed(123)
n <- 50; N <- 500
D_L <- rnorm(n, mean = 0.3)  # True difference
fD_L <- D_L + rnorm(n, sd = 0.5)
fD_U <- rnorm(N, mean = 0.3) + rnorm(N, sd = 0.5)
ppi_paired_test(D_L, fD_L, fD_U)
#> $estimate
#> [1] 0.2717196
#> 
#> $se
#> [1] 0.06818877
#> 
#> $z_stat
#> [1] 3.984814
#> 
#> $p_value
#> [1] 6.753292e-05
#> 
#> $reject
#> [1] TRUE
#> 
#> $lambda
#> [1] 0.741807
#> 
#> $ci
#> [1] 0.1380721 0.4053671
#>