Add noise to match outcomes — add_noise_to

This function can be used to flip a set proportion of match outcomes. Useful for testing models under varying levels of signal to noise ratios.

add_noise_to_outcomes(outcome, error_prob = 0.1)

Arguments

outcome	Original match outcome. Must be one of 0, 0.5, or 1.
error_prob	The probability an outcome switches from a 0 to 1 or vice-versa

Value

a numeric vector.

Details

Input outcomes of 0.5 will be randomy rounded to either 0 or 1.

Examples

# Create some sample outcomes
set.seed(1)
test_df <- data.frame(
   true_outcome = sample(c(0,1), 1000, replace=TRUE)
)

# No noise
test_df$noisy_outcome = add_noise_to_outcomes(test_df$true_outcome, error_prob = 0)
table(test_df)
#>             noisy_outcome
#> true_outcome   0   1
#>            0 502   0
#>            1   0 498

# Test different noise levels
test_df$noisy_outcome = add_noise_to_outcomes(test_df$true_outcome, error_prob = 0.1)
table(test_df)
#>             noisy_outcome
#> true_outcome   0   1
#>            0 450  52
#>            1  47 451

test_df$noisy_outcome = add_noise_to_outcomes(test_df$true_outcome, error_prob = 0.5)
table(test_df)
#>             noisy_outcome
#> true_outcome   0   1
#>            0 256 246
#>            1 228 270

test_df$noisy_outcome = add_noise_to_outcomes(test_df$true_outcome, error_prob = 1)
table(test_df)
#>             noisy_outcome
#> true_outcome   0   1
#>            0   0 502
#>            1 498   0