The University Pathogen Laboratory received blood samples from
25 patients (numbered 1, 2, etc.). Each blood sample was split into
two parts numbered 1A, 1B, 2A, 2B etc. Then all A- samples were
given to technician A, and all B- samples to technician B, who were
instructed to measure the HgA1c of the blood samples independently
of each another. The technicians’ results are given in the R codes
below. An analysis was carried out to test if there is a
significant difference in measurements between the two
technicians.
ID=seq(1:25)
A=c(6.8, 6.5, 6.2, 7.0, 6.5, 7.1, 6.9, 7.1, 6.0, 7.3, 6.4,
6.5, 6.0, 6.7, 5.4, 7.1, 7.1, 6.4, 5.9, 6.7, 7.6, 6.0, 6.3, 6.8,
6.3)
B=c(7.2, 6.7, 6.3, 7.5, 6.8, 7.5, 7.1, 7.2, 6.5, 7.6, 6.8, 6.7,
6.1, 7.2, 5.7, 7.5, 7.3, 6.5, 6.4, 7.0, 8.0, 6.2, 6.4, 7.3,
6.6)
t.test(A, B, paired=T)
This paired T-Test (done on R) is the same thing as a 1-Sample
T-Test with variable D=A-B. You can do it with a TI-84 calculator
using STAT->TESTS->T-Test.
But something went seriously wrong! You can see it by plotting
the differences B-A against the patient ID. What went wrong? How
could the Lab Director have prevented this wrongdoing (apart from
admonishing the technicians to behave ethically)?