Back to the Top
Hello:
Has anyone used the ordered categorical model in Trial Simulator? If
so, how does one get estimates for score values that separate each
pair of categories (ref. p. 202 of Examples guide) - this is in
reference to the Logit function.
Thanks in advance.
Best Regards
Dolly
__
Dolly A. Parasrampuria, PhD
Associate Director, Clinical Pharmacology
Drug Development & Safety,
M/C 263, McNeil Consumer & Specialty Pharmaceuticals, Inc.
7050 Camp Hill Rd, Fort Washington PA 19034-2299
Email: DParasr.-at-.mccus.jnj.com
Tel: 215-273-7760
Fax: 215-273-4127
Back to the Top
The following message was posted to: PharmPK
Dolly,
The estimates can be obtained from the plot on page 201 as the x values
corresponding to 50% on each of the three curves. The negatives of these
values are added to correct the order as required by the block
properties
disclaimer "Values should be in descending order." This, of course,
begs the
question, "How do you come up with the curves?" In the case of the
example,
my guess is that scores were simulated for a large number of subjects
and
the range of scores divided roughly into quartiles. Hope this is
helpful.
Cheers... Brian
Brian M. Sadler, PhD
Strategic PK Consulting, LLC
bsadler1.aaa.nc.rr.com
Back to the Top
The following message was posted to: PharmPK
Dear Dolly and Brian,
By definition, estimates for score values that separate each pair of
categories is the negative value of the input value that gives a 50%
probability of outputting either of the adjacent output values. This
is the
negative of the offset between outputs.
I haven't done the ordered categorical model in Trial Simulator, but
I think
these values should be obtained from some logit regression of the
real data
used to build the model. I agree with Brian that the curves can be
simulated
(or calculated) using the estimated parameters.
Thanks,
Qi
Qi_liu.aaa.merck.com
Back to the Top
The following message was posted to: PharmPK
Hi Qi,
You are always right to invoke real data and that would be my choice in
practice. I was merely providing my best guess as to how the values
used in
the Trial Simulator example were derived. Perhaps the two other TS2
instructors, Mark Lovern (now at GSK) or Peter Vis (now at Exprimo), can
provide more details on the example in question.
Cheers... Brian
Brian M. Sadler
Strategic PK Consulting, LLC
bsadler1.-a-.nc.rr.com
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)