Top-left statistic

Calculate the optimal probability classification threshold using the point closest to the top-left corner of the ROC plot.

Usage

calc_topleft(estimates, truth)

Arguments

estimates

A vector of classification probabilities. Values should represent the probability of 1 in the truth argument.

truth

An integer vector of 0 and 1 representing the true classifications.

Value

A numeric scalar representing the optimal probability threshold.

Details

The top-left corner of an ROC plot represents perfect sensitivity and specificity. The top-left statistic is determined by calculating the distance from each point on the ROC curve (e.g., create_roc()) to the top-left corner of the plot. The point on the plot that is the closest to the top-left corner is the optimal threshold. The optimality criterion is then defined as:

$$\text{min}((1 - sensitivity)^2 + (1 - specificity)^2)$$

See also

Other optimal threshold methods: calc_cz(), calc_gmean(), calc_youden()

Examples

calc_topleft(estimates = dcm_probs$att1$estimate,
             truth = dcm_probs$att1$truth)
#> [1] 0.3170266

calc_topleft(estimates = dcm_probs$att2$estimate,
             truth = dcm_probs$att2$truth)
#> [1] 0.4363361

calc_topleft(estimates = dcm_probs$att3$estimate,
             truth = dcm_probs$att3$truth)
#> [1] 0.3722605