Skip to contents

A comprehensive table containing all pairwise comparisons between 68 diagnostic algorithms for Parkinson’s Disease Dementia (PDD) evaluated in the study. The table serves as the basis for Figures 2, A3, A4, and A5, as well as for other summary results presented throughout the article.

Usage

concords

Format

A tibble with 7,744 rows and 21 columns:

predictor

Character; name of the algorithm used as a predictor.

reference

Character; name of the algorithm used as a reference.

N

Integer; the number of patients used to do calculations.

Kappa

Character; Cohen’s kappa with its 95% confidence interval.

Accuracy

Character; accuracy estimate with its 95% confidence interval.

Kappa_raw

Numeric; Cohen’s kappa estimate.

McnemarPValue

Numeric; p-value from McNemar’s chi-squared test for symmetry between rows and columns.

Accuracy_raw

Numeric; accuracy estimate.

NoInformationRate_raw

Numeric; no-information rate, i.e., the accuracy if all patients were classified as non-PDD.

AccuracyPValue

Numeric; p-value from the Exact Binomial Test comparing accuracy to the no-information rate.

Sensitivity

Numeric; sensitivity estimate.

Specificity

Numeric; specificity estimate.

Pos Pred Value

Numeric; positive predictive value.

Neg Pred Value

Numeric; negative predictive value.

Precision

Numeric; precision estimate.

Recall

Numeric; recall estimate.

F1

Numeric; F1 score.

Prevalence

Numeric; PDD prevalence according to the reference algorithm.

Detection Rate

Numeric; detection rate.

Detection Prevalence

Numeric; detection prevalence.

Balanced Accuracy

Numeric; balanced accuracy, i.e., (sensitivity + specificity) / 2.

Source

Generated by the targets pipeline in _targets.R.

Details

Each algorithm is used once as a reference and once as a predictor, meaning that each algorithm pair appears twice. Some metrics, such as Cohen’s kappa, are symmetrical, whereas others (e.g., accuracy, sensitivity, or specificity) are not.

See also