Comparison of traditional, trimmed traditional and robust Youden charts

Traditional Youden chart can be distorted by non-normally distributed data. Here, an optimized Youden chart was developed and compared with the traditional and trimmed traditional Youden charts.

The urea concentrations were determined by 28 laboratories to provide data for the construction of Youden charts. Normality of these data was tested. Outliers were excluded prior to construction of the trimmed traditional Youden chart. Non-robust and robust estimators were computed to construct the traditional and robust Youden charts, respectively. Robust between-laboratory z-score (ZBi) and within-laboratory z-score (ZWi) were obtained to assess whether or not these charts can reasonably present the urea results. Expected outcomes were the points related to acceptable (|ZBi| and |ZWi| ≤ 2), questionable (2 < |ZBi| < 3 and/or 2 < |ZWi| < 3), and unacceptable (|ZBi| and/or |ZWi| ≥ 3) results fall inside, on/near, and outside the ellipse, respectively.

Only the data from lot 201111 are non-normally distributed. Five- and 2-pair outliers are excluded from the data of lots 201111 and 201112, respectively. The concordance rates of the traditional, trimmed traditional and robust Youden charts are 87.1%, 92.9%, and 94.3%, respectively.

Among the three charts, the robust Youden chart presents the urea results best.