DTA SROC Explorer

Study data

Data CSV: label, TP, FP, FN, TN

TP = true positives, FP = false positives, FN = false negatives, TN = true negatives. With a covariate, add it as a last column. Example data (datasets shipped with the R package mada) Choose a dataset, then press Load example.

Bivariate model Estimation method Continuity correction (0.5) for zero cells Meta-regression A numeric covariate enters as it is; text values are treated as categories, with the alphabetically first category as reference (as in R).

Display Plot title

The ROC plot uses the fixed [0,1] domain on both axes. Axis controls are disabled.

ROC space + Moses & bivariate SROC

Fagan nomogram (bivariate LR±)

Deeks' funnel (small-study bias)

Bivariate model (Reitsma et al. 2005): logit sensitivity and logit false-positive rate are pooled jointly with a between-study covariance matrix, by the estimators of the R package mada (function reitsma, which calls mvmeta): REML (default), ML, method of moments, variance components, or fixed effect; with an optional study-level covariate (bivariate meta-regression). Shown in crimson: the summary point (♦), the Rutter–Gatsonis SROC curve, and the 95% confidence (solid) and prediction (dashed) regions, as plot.reitsma(predict = TRUE). Computed in pure JavaScript and checked against mada 0.5.12 on all six of its example datasets. Moses–Littenberg SROC (blue; Moses et al., Stat Med 1993): unweighted OLS of D = logit(Se) − logit(FPR) on S = logit(Se) + logit(FPR), as mada::mslSROC; it ignores between-study heterogeneity and is shown for comparison. Related: the HSROC explorer → fits the Rutter & Gatsonis (2001) parameterisation directly.