fyy897854960
☆    

China,
2018-11-01 10:28
(1974 d 12:37 ago)

(edited by mittyri on 2018-11-01 11:53)
Posting: # 19512
Views: 5,501
 

 About the Partial SS in Boequivalence results in WinNonlin 8.1 [Software]

Dear All,
I have some questions about the Partial SS in Bioequivalence results in WinNonlin.

Firstly, why the Partial SS results just provided under default model(2x2x2 design) in WinNonlin 8.1? It just provides the Partial test worksheet in other models. What are the differences and relationship between the Partial test results and Partial SS results?

Secondly, what are the Numer_DF and Denom_DF? What is the calculation method for these DF since it could be decimals in Degrees of freedom sometimes. Just like the following results:

Dependent Units Hypothesis Numer_DF Denom_DF F_stat P_value
Ln(AUClast) int 1 17.863275 2.012866 0.17318289
Ln(AUClast) Sequence 1 17.876226 9.2107972 0.007163829
Ln(AUClast) Formulation 1 17.189233 0.19446307 0.66472393
Ln(AUClast) Period 1 17.202333 0.67862197 0.42133666

Could you kindky help me with these questions?

Thank you very much,

Fu Yangyang


Edit: Category changed; see also this post #1. [Mittyri]
mittyri
★★  

Russia,
2018-11-01 12:50
(1974 d 10:14 ago)

@ fyy897854960
Posting: # 19513
Views: 4,838
 

 About the Partial SS in Boequivalence results in WinNonlin 8.1

Dear Fu Yangyang,

❝ Firstly, why the Partial SS results just provided under default model(2x2x2 design) in WinNonlin 8.1?

Note that General Linear Model is applicable for a model without additional variance structures (PURE ANOVA). Partial SS is very similar to SS type III table provided by SAS (GLM procedure)

❝ It just provides the Partial test worksheet in other models. What are the differences and relationship between the Partial test results and Partial SS results?

RTFM (WNL User's guide)!
Partial Tests worksheet
The Partial Tests worksheet is created by testing each model term given every other model term.
Unlike sequential tests, partial tests are invariant under the order in which model terms are listed in the Fixed Effects tab. Partial tests factor out of each model term the contribution attributable to the remaining model terms.
This is computed by modifying the basis created by the QR factorization to yield a basis that more closely resembles that found in balanced data.
Sequential SS and Partial SS worksheets
For models with only fixed effects, the Sequential Tests and Partial Tests are also presented in a form that includes SS (Sum of Squares) and MS (Mean Square).


❝ Secondly, what are the Numer_DF and Denom_DF? What is the calculation method for these DF since it could be decimals in Degrees of freedom sometimes.

Same source:
The default calculation method of the degrees of freedom is controlled on the General Options tab and is initially set to Satterthwaite.
With that method of DF approximation float DFs are likely, especially in case of some inbalance.

PS: Aren't you satisfied with Certara Support Forum?:-D

Kind regards,
Mittyri
fyy897854960
☆    

China,
2018-11-01 15:16
(1974 d 07:49 ago)

(edited by mittyri on 2018-11-01 16:26)
@ mittyri
Posting: # 19517
Views: 4,725
 

 About the Partial SS in Boequivalence results in WinNonlin 8.1

❝ PS: Aren't you satisfied with Certara Support Forum?:-D

Sometimes I can not get timely reply. So I post my question here.

Anyway, thanks again.

Fu


Edit: Full quote removed. Please delete everything from the text of the original poster which is not necessary in understanding your answer; see also this post #5[Mittyri]
ElMaestro
★★★

Denmark,
2018-11-01 23:49
(1973 d 23:15 ago)

@ fyy897854960
Posting: # 19518
Views: 4,748
 

 About the Partial SS in Boequivalence results in WinNonlin 8.1

Hi Fyy et al.,

I am not a SAS user or a WNL user, but I think the basis for this question may be a little wrong?

In a 222BE trial the result is generally produced with a linear model, not a mixed model, regardless of authority. In that situation the whole concept of denominator df's if irrelevant. The residual df's will be an integer.

So, I am guessing you fit your model with subject as random, and hence it becomes a mixed model.
If I recall correctly WNL happily fits subject as random by default so one may need to actively tell the beast that everything including subject is fixed. Then the analysis becomes compliant with guidelines and expectation, and there will be no denominator df's to confuse you. :-)

Pass or fail!
ElMaestro
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2018-11-02 00:12
(1973 d 22:53 ago)

@ ElMaestro
Posting: # 19519
Views: 4,623
 

 RTFM

Hi ElMaestro et al.,

❝ In a 222BE trial the result is generally produced with a linear model, not a mixed model, regardless of authority. […]


❝ If I recall correctly WNL happily fits subject as random by default so one may need to actively tell the beast that everything including subject is fixed.


Yep. It is possible to set that in the preferences of Phoenix for a good while (too lazy to search since when). Then any study will be evaluated with fixed effects only.

@Fu Yangyang: RTFM.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
mittyri
★★  

Russia,
2018-11-02 12:19
(1973 d 10:45 ago)

@ ElMaestro
Posting: # 19524
Views: 4,764
 

 Hidden features of WNL Bioequivalence

Hi ElMaestro et al.,

❝ In a 222BE trial the result is generally produced with a linear model, not a mixed model, regardless of authority. In that situation the whole concept of denominator df's if irrelevant. The residual df's will be an integer.

Yes, that's correct. I think the second topicstarter's question is related to replicate designs.

❝ So, I am guessing you fit your model with subject as random, and hence it becomes a mixed model.

In SAS (and R) the user is free to to use different approaches with different procedures (i.e. PROC GLM, PROC MIXED ...).
In WNL you have only one module which is MIXED by default.
But what about RANDOM statements inside PROC GLM (remember the usual FDA SAS code for sinple 2X2 crossover)?
See the guide:
Fixed effects model: Sequence + Formulation + Period
– Unless the bioequivalence preference Default for 2x2 crossover set to all fixed effects is turned on in the Preferences dialog (Edit > Preferences > LinMixBioequivalence), in which case the model is  Sequence + Subject(Sequence) + Formulation + Period.
Random effects model: Subject(Sequence) and Type: Variance Components
– Unless the bioequivalence preference Default for 2x2 crossover set to all fixed effects is turned on, in which case the model is not specified (the field is empty).

But is it MIXED or GLM+RANDOM?
Later:
Since there is no repeated specification, the default error model ε ~ N(0, σ2 I) is used. This is equivalent to the classical analysis method, but using maximum likelihood instead of method of moments to estimate inter-subject variance. Using Subject as a random effect this way, the correct standard errors will be computed for sequence means and tests of sequence effects.
Here ya go! This is GLM with RANDOM!

That's important hidden feature:
- the design (replicate/non-replicate) is caught on the fly,

- for 2X2 bioequivalence fixed(Sequence + Formulation + Period), Subject as Random mean GLM+RANDOM,

- for replicated designs the default model is the model proposed by FDA (also EMA Method C):
Fixed effects model: Sequence + Formulation + Period
Random effects model: Subject(Sequence) and Type: Variance Components
Repeated specification: Period
– Variance Blocking Variables: Subject
– Group: Treatment
– Type: Variance Components

And you need to manually change the model to use EMA ABEL (Method A or B)

- for replicated designs fixed(Sequence + Formulation + Period) and Subject as Random mean MIXED model (and you don't get SS, DFs could be floats). This is EMA Method B.

- To switch to ANOVA with replicated designs you need to manually move subject to fixed terms tab and remove everything from Random Tab.


PS: all these things Simon and other trainers are explaining during training sessions ;-)

Kind regards,
Mittyri
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
65 visitors (0 registered, 65 guests [including 12 identified bots]).
Forum time: 23:05 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5