Summary
A robust two-way analysis of variance technique was applied to determine simultaneously the effects of method and thromboplastin on prothrombin time. A new approach to outlier detection for two-way analysis of variance was used. Focusing on the underlying error structure improved the uniformity of the grading procedure in the hematology proficiency testing program of the New York State Department of Health.
The logarithm-transformed scale produced constancy of error variance and resulted in uniformity of the acceptable spread of data. The common variance was lower than that obtained by previous methods and allowed for a narrower acceptable range of reported prothrombin times by reducing the inflated standard deviation, thus improving the efficiency of the grading procedure.
For proficiency testing, no advantage was found in the use of either a common thromboplastin or freeze-dried, coumadinized patient plasmas rather than artificially depleted commercial plasmas, except for special purposes.
Keywords
Proficiency testing - Robust estimation - Outliers - Prothrombin time - Instrumentation - thromboplastin variables
