Testing For Total Error
•
U
estimated by combining random and systematic error.
•
Method Comparisons:
•
U
40 - 100 patient specimens are tested by running:
•
e
on a known accurate method and current
method
•
e
on a current method and new method
•
U
perform statistical analysis on data:
•
e
t-test
•
e
Correlation coefficient
•
0 = no correlation
•
+1 = perfect correlation
•
e
Linear regression