|
This talk will focus on two issues in method comparison studies. First,
if there is "sufficient" agreement between two competing instruments,
the less expensive or the easier-to-use instrument is preferable. The
statistical problem here is to assess the degree of agreement. An
overview of the existing practices will be presented, followed by three
new approaches for testing the hypotheses of satisfactory agreement.
The appropriate approach to use is guided by goals of the experiment
and the experimenter's ability to specify the thresholds. The
advantages of the new approaches over the existing ones will be
illustrated using a dataset from medical literature. Second, if two
instruments are compared with a gold-standard, it is desirable to
select the instrument that agrees most with the gold-standard. A large
sample two-stage indifference zone procedure will be presented for this
selection problem. Questions like which parameterization is the "best"
and what first-stage sample size is "adequate" are answered using
simulation. This technique will also be illustrated using a dataset
from literature.
(This is joint work with my advisor Prof. H. N. Nagaraja.) |
| Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served. |