Close  
Figure 2: These plots show the diagnostic rates in relation to the group median (zero). Positive numbers imply a relative over call and negative numbers imply a relative under call. As the numbers are generated by dividing through by the standard error, pathologists that read different numbers of cases can readily be compared. Each plot represents one pathologist. Diagnoses that have a standard error > 2 or <−2 are unlikely due to sampling (P < 0.05). Diagnoses that have a standard error >3 or <−3 are very unlikely (P < 0.001) to be due to sampling alone, and suggest diagnostic bias. (a) Pathologist 4, (b) pathologist 9, (c) pathologist 10, (d) pathologist 11, (e) pathologist 13, (f) pathologist 14, (g) pathologist 17, (h) pathologist 20 
