Log in
RegisterBogus PROC MIXED/FDA [Software]
Dear dixit,
I fully agree with EM.
IMHO Proc MIXED stops here with arbitrary values for the inter-inividual variance parameters and also with arbitrary intra-individual variability for Test. An indication is:
❝ 2 treatment T 1 0.4233 0.5232 CVinter=82.9%
❝
❝ Covariance Parameter Estimates
Do you really believe that your Test preparation has such a very lower intra-individual variability? If it is the same drug I can't imagine that at all
.
Once again said, the FDA model is over-specified for the partial replicate design. And therefore the model fit with Proc MIXED is not reliable.
We had this phenomenon already here were SAS Proc MIXED and WINNONLIN gave totally different values for the covariance parameters.
What to the rescue?
Within a model formulation in Proc MIXED I don't know, as I already said. I have the strong believe (but believe is not The power to know
) that it is impossible to get a solution within that because we have a confounding between the subject-by-formulation interaction (an inter-individual term) and the intra-individual variability of the Reference in the design used here.
I would suggest you to go with the so called "Methods of moments".
See
[1]R.J. McNally
Tests for Individual and Population Bioequivalence Using 3-Period Crossover Designs
which can be found here for the necessary statistics within a partial replicate design.
They can be used also in the framework of ABE and scaled ABE. Your intention I guess?
Or neglect any subject-by-formulation interaction and go with the classical model (same as for a 2x2 cross-over) to obtain the ABE 90% confidence intervals and estimate independently the intra-individual variability of the Reference using the intra-subject contrasts
Good luck.
BTW: Seems no m on your keyboard
.
I fully agree with EM.
IMHO Proc MIXED stops here with arbitrary values for the inter-inividual variance parameters and also with arbitrary intra-individual variability for Test. An indication is:
❝ Estimated G Matrix
❝
❝ Row Effect treatment subject Col1 Col2
❝ 2 treatment T 1 0.4233 0.5232 CVinter=82.9%
❝
❝ Covariance Parameter Estimates
❝
❝ Cov Parm Subject Group Estimate
❝ ...
❝ Residual subject treatment R 0.6289 CVintra=93.6%
❝ Residual subject treatment T 0.09081 CVintra=30.8%
❝
Do you really believe that your Test preparation has such a very lower intra-individual variability? If it is the same drug I can't imagine that at all
.Once again said, the FDA model is over-specified for the partial replicate design. And therefore the model fit with Proc MIXED is not reliable.
We had this phenomenon already here were SAS Proc MIXED and WINNONLIN gave totally different values for the covariance parameters.
What to the rescue?
Within a model formulation in Proc MIXED I don't know, as I already said. I have the strong believe (but believe is not The power to know
) that it is impossible to get a solution within that because we have a confounding between the subject-by-formulation interaction (an inter-individual term) and the intra-individual variability of the Reference in the design used here.I would suggest you to go with the so called "Methods of moments".
See
[1]R.J. McNally
Tests for Individual and Population Bioequivalence Using 3-Period Crossover Designs
which can be found here for the necessary statistics within a partial replicate design.
They can be used also in the framework of ABE and scaled ABE. Your intention I guess?
Or neglect any subject-by-formulation interaction and go with the classical model (same as for a 2x2 cross-over) to obtain the ABE 90% confidence intervals and estimate independently the intra-individual variability of the Reference using the intra-subject contrasts
(R-R')/sqrt(2).Good luck.
BTW: Seems no m on your keyboard
.—
Regards,
Detlew
Regards,
Detlew
Complete thread:
- Convergence problem in PROC MIXED dixit 2010-03-21 10:19
![[Mix]](https://static.bebac.at/img/mix.png "open in Mix view")
- Convergence problem in PROC MIXED ElMaestro 2010-03-21 11:35
- Convergence problem in PROC MIXED dixit 2010-03-22 06:26
- Convergence problem in PROC MIXED ElMaestro 2010-03-22 08:21
- Convergence problem in PROC MIXED dixit 2010-03-23 19:15
- Convergence problem in PROC MIXED ElMaestro 2010-03-23 20:29
- Convergence problem in PROC MIXED dixit 2010-03-24 15:09
- Convergence problem in PROC MIXED ElMaestro 2010-03-25 07:44
- Bogus PROC MIXED/FDAd_labes 2010-03-25 09:15
- Bogus PROC MIXED/FDA ElMaestro 2010-03-25 10:42
- Bogus PROC MIXED/FDA d_labes 2010-03-26 09:45
- Bogus PROC MIXED/FDA ElMaestro 2010-03-25 10:42
- Convergence problem in PROC MIXED dixit 2010-03-24 15:09
- Convergence problem in PROC MIXED shri 2010-04-05 08:01
- Convergence problem in PROC MIXED ElMaestro 2010-03-23 20:29
- Convergence problem in PROC MIXED dixit 2010-03-23 19:15
- partial replicate? d_labes 2010-03-22 09:50
- Convergence problem in PROC MIXED ElMaestro 2010-03-22 08:21
- Convergence problem in PROC MIXED dixit 2010-03-22 06:26
- Convergence problem in PROC MIXED ElMaestro 2010-03-21 11:35
Ing. Helmut Schütz