Diagonal covariance structure, un-unique intra subject levels

[PharmCat]

Summary

[danielinteractive]

Hi @PharmCat , can you please add details?

[PharmCat]

Diagonal covariance structure and un-unique intra-subject levels. Solve some models where grouped repeated observations have factorized uncorrelated covariance structure.

• Provide examples of real-world use cases that this would enable
  and how it solves the problem you described.

Bioequivalnce study - to estimate intra-subject variation for different drugs.

• How do you solve this now?

SPSS, SAS, Julia (Metida)

• Have you tried to work around the problem using other tools?

For R I did'nt found solution.

I think that in case of diagonal (uncorrelated) structure requirements for levels uniqueness can be omitted.

[clarkliming]

it sounds like #174

[danielinteractive]

Or #300 ?

[PharmCat]

HI! Yes it's related to In the FDA guidance for bioequivalence studies. For BE study SAS can just group repeated effect without limitation and we get variance for both treatment separately:

PROC MIXED data=pkdata;
CLASSES Seq Subject Period Formulation;
MODEL  logData = Seq Period Formulation/ DDFM=SATTERTH;
RANDOM  Formulation/TYPE=FA0(2) SUB=Subject G;
REPEATED/GRP=Formulation SUB=Subject;
ESTIMATE 'T vs. R' Formulation -1 1/CL ALPHA=0.1;
RUN;

In SPSS we have similar problem - cant fit with un-qnique repeated levels, but there is workaround

MIXED logData BY Period Formulation Sequence Subject
  /CRITERIA=CIN(90) MXITER(200) MXSTEP(20) SCORING(2) SINGULAR(0.000000000001) HCONVERGE(0,
    RELATIVE) LCONVERGE(0.0000000000001, RELATIVE) PCONVERGE(0, RELATIVE)
  /FIXED=Period Formulation Sequence | SSTYPE(3)
  /METHOD=REML
  /RANDOM=Formulation | SUBJECT(Subject) COVTYPE(CSH)
  /REPEATED=Formulation | SUBJECT(Subject*Period) COVTYPE(DIAG)
  /EMMEANS=TABLES(Formulation) COMPARE REFCAT(FIRST) ADJ(LSD).

/REPEATED=Formulation | SUBJECT(Subject*Period) COVTYPE(DIAG)

And this ask only about repeated effects :) not about random (but if you add random part - if will be great :) )