Bioequivalence and Bioavailability Forum

randombadger
☆

UK,
2011-06-22 13:45
(5472 d 09:53 ago)

Posting: # 7158
Views: 8,486

Estimated G matrix not + definite when using FDA model [General Statistics]

Hi,

I have a 3 period, 2 treatment crossover (TRR, RTT) design and used the following code (as per FDA guidance) to analyse the data:

PROC MIXED DATA= cmax ;

  CLASSES armcd usubjid period treat ;

  MODEL LOG_PARM = armcd period treat / DDFM=KR;

  RANDOM treat/TYPE=FA0(2) SUB=usubjid G;

  REPEATED/GRP=treat SUB=usubjid;

  ESTIMATE 'B vs A' treat -1 1/CL ALPHA=0.1;

  LSMEANS treat/ cl alpha=0.1;

RUN;

The above analysis resulted in the following note in the SAS log:
"Estimated G matrix not positive definite"
suggesting that "one or more variance components on the RANDOM statement is/are estimated to be zero and could/should be removed from the model."

My colleague is suggesting I switch the covariance structure to VC however I'm concerned that this negates the REPEATED statement and is questionable in a 3x2 design. What is your opinion on this point?

I used the CSH structure but, as expected, encountered the same notes in the log.

Looking through the covariance structures, Compound Symmetry (CS) looks more appropriate as it accounts for correlation & assumes it is constant regardless of the lag between pairs of repeated measurements (sounds OK given the study design).

Any opinions on the above points gratefully received.
Thanks,
RB

Edit: Formatted. You may use BBCodes (see here). [Helmut]

Helmut
★★★

Vienna, Austria,
2011-06-22 15:36
(5472 d 08:01 ago)

@ randombadger
Posting: # 7159
Views: 7,303

overspecified model?

Post reply

Dear RB!

❝ I have a 3 period, 2 treatment crossover (TRR, RTT) design and used the following code (as per FDA guidance) to analyse the data:

❝

❝ FDA 2001 Apendix E

❝

❝ The above analysis resulted in the following note in the SAS log: "Estimated G matrix not positive definite" suggesting that "one or more variance components on the RANDOM statement is/are estimated to be zero and could/should be removed from the model."

Yes, old story. SAS complains about an over-specified model (T is not repeated) and gives a variance for T. ;-)

See this thread.

❝ My colleague is suggesting I switch the covariance structure to VC however I'm concerned that this negates the REPEATED statement and is questionable in a 3x2 design. What is your opinion on this point?

Agree.

❝ I used the CSH structure but, as expected, encountered the same notes in the log.

Yep.

❝ Looking through the covariance structures, Compound Symmetry (CS) looks more appropriate as it accounts for correlation & assumes it is constant regardless of the lag between pairs of repeated measurements (sounds OK given the study design).

Interesting idea. At least you don’t get a warning any more. :-D

Tried it with EMA’s example data set II (TRR|RTR|RRT) and got (Phoenix 6.2 = poor man’s SAS):

                         90% CI         CV_WR   CV_WT

EMA's Model A         97.32  107.46

EMA’s Model B         97.32  107.46

EMA’s crippled model       NA          11.20   NA

FDA FA0(2)            97.05  107.76    11.55   8.65 Warning (SAS: CV_WT=3.87)

FDA CSH               not estimable     8.26   7.15 Warning

FDA CS                97.05  107.76    11.55  18.67 no Warning!

Pro: CV_WR is identical to FA0(2) and you don’t get a warning with CS.
Con: That’s not what FDA expects you to do: FA0(2) or possibly [sic!] CSH. I have strong doubts whether the value of CV_WT makes any sense.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

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

d_labes
★★★

Berlin, Germany,
2011-06-29 11:39
(5465 d 11:58 ago)

@ Helmut
Posting: # 7184
Views: 7,062

Overspecified model

Post reply

Dear Helmut!

❝ ... Yes, old story. SAS complains about an over-specified model (T is not repeated) and gives a variance for T ...

Here you are in mistake. Randombadgers sequences are TRR / RTT, R replicated in the first sequence group, T replicated in the second.

Nevertheless the model may be over-specified if the subject-by-treatment interaction in the model turns out to be zero for the analysed data set.

The CS (compound symmetric) structure is then the more appropriate but is also to some extent over-specified. See my attempt to dance.

—
Regards,

Detlew

Helmut
★★★

Vienna, Austria,
2011-06-29 16:11
(5465 d 07:26 ago)

@ d_labes
Posting: # 7192
Views: 6,982

Overspecified model

Post reply

Dear D. Labes!

❝ ❝ ... Yes, old story. SAS complains about an over-specified model (T is not repeated) and gives a variance for T ...

❝

❝ Here you are in mistake. Randombadgers sequences are TRR / RTT, R replicated in the first sequence group, T replicated in the second.

Oops; you are right, of course. Need a vacation.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

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

d_labes ★★★ Berlin, Germany, 2011-06-29 17:45 (5465 d 05:53 ago) @ Helmut Posting: # 7193 Views: 6,920	To bike Post reply
	Dear Helmut! ❝ Oops; you are right, of course. Need a vacation. I can warmly recommend you a bike tour with 'Dutch bike tours' across The Netherlands. According to my very recent experience you will get all you need for an impressive and unforgettable vacation: Stormy wind from ahead, heavy rain, cold temperatures, hotels directly beside highways. — Regards, Detlew

Helmut
★★★

Vienna, Austria,
2011-06-29 18:03
(5465 d 05:35 ago)

@ d_labes
Posting: # 7194
Views: 6,938

To bike

Post reply

Dear D. Labes!

❝ I can warmly recommend you a bike tour with 'Dutch bike tours' across The Netherlands.

Well, I learned to ride a bike on my 30^th birthday (which is quite a while ago). Practiced ever since only a couple of times. ;-)

❝ According to my very recent experience you will get all you need for an impressive and unforgettable vacation:

❝ Stormy wind from ahead, heavy rain, cold temperatures, hotels directly beside highways. :cool:

I thought while biking you always suffer from headwinds (Murphy’s law #67)?

My Hotel; panoramic view from the balcony …
[image]

… and from back (so far about highways and the hidden beauty of architectonic ensembles):
[image]

Puzzle
[image]

Didn’t succeed in figuring out how to switch off the seat’s heating. Another five buttons, LEDs in different colors, no icons, lot of text – in Japanese only. Considered trial-and-error far too risky.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

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

Estimated G matrix not + definite when using FDA model [General Sta­tis­tics]

overspecified model?

Overspecified model

Overspecified model

To bike

To bike

Estimated G matrix not + definite when using FDA model [General Statistics]