## Nonlinear Mixed Effect Model Analysis

Occasional Contributor
Posts: 10

# Nonlinear Mixed Effect Model Analysis

Hello.. I am doing analysis of repeated measure data obtained from a clinical trial. There are total 12 time points (one baseline and 11 follow-up) and time is coded in Hours with baseline time coded as "0".  Outcomes of trial are three, one continuous (pain assessment using 100 mm VAS scale) and two binary (both binary outcomes coded as 0 and 1). Two predictors (covariates) are sex (male coded as 0 and female coded as 1) and age groups (pre-adolescent coded as 0 and post-adolescent coded as 1). Therefore  from statistical analysis point of view, it is a 2x2 factorial design. Objective of my trial are to evaluate the main effect and interaction effect of age and sex on pain (VAS score). The mean profile plot shows a nonlinear quadratic trend of VAS score.  I am familiar with PROC MIXED AND PROC GLIMMIX procedure and one possibility is to use polynomial random coefficient model analysis using these two aforementioned procedures. But my objective is not to investigate the trend of pain but to evaluate the overall effect of age, sex and age sex interaction on VAS score. Therefore, i was just thinking about the use of PROC NLMIXED in this case. But honestly speaking, i find it very difficult to use NLMIXED procedure and not able to understand syntax and its application. for example, how to get initial PARMS values etc etc. PROC GLIMMIX is very versatile procedure and i am using it for my two binary outcomes. Is there any possibility that i can use PROC GLIMMIX for nonlinear longitudinal data analysis instead of using PROC NLMIXED?

Regards

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

When I think of nonlinear, I assume it to mean nonlinear in the parameters, such that a polynomial is not what I think of as nonlinear.  Given that, it makes perfect sense to me to use GLIMMIX to explore this.  If you have access to SAS/STAT 12.1, check out the EFFECT statement.  It will enable you to examine polynomial effects specifically (as well as spline effects, which may be more appropriate).

Steve Denham

Occasional Contributor
Posts: 10

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Dear Steve Denham,

Thanks a lot for replying. I am using SAS 9.2 and do not have access to SAS/STAT 12.1. Following is the code that i can think of fitting a second order (quadratic) polynomial trend by using GLIMMIX

proc glimmix data=data ORDER=Data nobound;

class id sex age;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time

/ ddfm=kr s cl ;

random int time time*time / subject=id type=un v;

lsmeans sex / pdiff adjust=simulate(acc=.0005 seed=121211 report) adjdfe=row cl;

lsmeans age / pdiff adjust=simulate(acc=.0005 seed=121211 report) adjdfe=row cl;

lsmeans sex*age / pdiff adjust=simulate(acc=.0005 seed=121211 report) adjdfe=row cl;

run;

is this code appropriate for analysis of quadratic trend? should i use random _residual_ statement also to model covariance structure e.g.

random _residual / sub=id type=sp(pow)(xtime). i am thinking of spatial power because my time points are not equally spaced. i can use xtime as class variable and time as continuous variable.  So code will be like this:

proc glimmix data=data ORDER=Data nobound;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time

/ ddfm=kr s cl ;

random int time time*time / subject=id type=un v;

random _residual / sub=id type=sp(pow)(xtime);

lsmeans sex / pdiff adjust=simulate(acc=.0005 seed=121211 report) adjdfe=row cl;

lsmeans age / pdiff adjust=simulate(acc=.0005 seed=121211 report) adjdfe=row cl;

lsmeans sex*age / pdiff adjust=simulate(acc=.0005 seed=121211 report) adjdfe=row cl;

run;

is this code for analysis and multiple comparison by using lsmeans statement appropriate? or should i use ESTIMATE statement for doing multiple comparisons.

Thanks a lot for the help

Regards

Sandhu

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

This looks exactly like what you will need.  I will be curious to find out if the random _residual_ statement will work.  I think it should, but I worry some that if you have a large number of xtime levels, you may have convergence problems.

Good luck, and please let the community know if this approach worked.

Steve Denham

Occasional Contributor
Posts: 10

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Dear Steve Denham,

I tried this code with random _residual_ statement for my data which comprised of 280 subject and each subject had 12 repeated measurements. your prediction was correct. It did not work. The log was :

NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.

WARNING: Optimization stopped because of infinite objective function.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           1.01 seconds

cpu time            0.96 seconds

It converged but "Optimization stopped because of infinite objective function." this happened with both Dual Quasi-Newton and  Newton-Raphson with Ridging optimization techniques. If i am correct, the  Newton-Raphson with Ridging technique makes PROC GLIMMIX almost similar to PROC MIXED. So this mean random _residual_  in PROC GLIMMIX or Repeated statement Repeated xtime/ subject=id type=sp (pow)(xtime) in PROC MIXED will not work? if so, then how to model covariance structure.

Regards

Sandhu

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

If the model converged but the optimization stopped, I would look at the ddfm=kr option on the model statement as the most likely suspect, as this is a secondary process applied to the converged model. I am pretty sure that the unequal spacing leads to some problems with the full Kenward-Rogers procedure under unbalanced situations. You might be able to work around this with ddfm=kr(firstorder).  If your data are balanced across sex and age, then you could try ddfm=satterthwaite.  And if there are still problems, ddfm=bw, where bw is between-within.

Steve Denham

Occasional Contributor
Posts: 10

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Dear Steve Denham,

There are couple of things which i feel are responsible for these problems.

When i was running the original quadratic code, i was getting G matrix not positive warning as follow:

proc glimmix data=data ORDER=Data ;

class id sex age ;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

run;

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           22.37 seconds

cpu time            22.12 seconds

Therefore, i used nobound option in the procedure as recommended (by Little 2006) for PROC MIXED procedure:

proc glimmix data=data ORDER=Data nobound ;

class id sex age ;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

run;

It worked fine for quadrartic code even after adding nobound.

Now when i added random _residual_ statement in this above code along with nobound option, the problem of "Optimization stopped because of infinite objective function." as i mentioned in the previous post. This problem persisted even with ddfm=kr(firstorder),   ddfm=satterthwaite and ddfm=bw .

But when i removed nobound, it worked even with ddfm=kr.

proc glimmix data=data2 ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

random _residual_ / sub=id type=sp(pow)(xtime);

run;

Log is:

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           23.30 seconds

cpu time            23.10 seconds

Now here is my dilemma : should i use nobound option and no random _residual_ statement OR should i use random _residual_ statement and no nobound option. IF i am not using nobound option, then how to justify the effect of Estimated G matrix is not positive definite.

These above codes were with default Optimization Technique i.e.  Dual Quasi-Newton.

Now part two of the problem is changing  Optimization Technique from  Dual Quasi-Newton to Newton-Raphson with Ridging with nloptions tech= nrridge.

when i run same quadratic code (without random _residual_ statement) along with nobound option as :

proc glimmix data=data2 ORDER=Data  nobound;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

nloptions tech= nrridge;                *Newton-Raphson with Ridging technique*;

run;

there is no difference in results for from previous Optimization Technique i.e  Dual Quasi-Newton. The Fit criteria, Type III test of fixed effect are exactly same, so as the :

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           3.75 seconds

cpu time            3.71 seconds

This Newton-Raphson with Ridging technique also did not work with nobound option as it happened with  Dual Quasi-Newton after i added random _residual_ statement.

BUT The problem occurs when i use nloptions tech= nrridge (i.e Newton-Raphson with Ridging technique) along with random _ residual_ statement but without nobound option.

proc glimmix data=data ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

random _residual_ / sub=id type=sp(pow)(xtime);

nloptions tech= nrridge;

run;

now log is:

NOTE: Convergence criterion (XCONV=0) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           10.43 seconds

cpu time            10.34 seconds

IN CONTRAST to log from same statement but with Dual Quasi-Newton technique, the above log (with Newton-Raphson with Ridging technique) shows No NOTE  of Estimated G matrix is not positive definite and  Convergence criterion changes from (GCONV=1E-8) to  (XCONV=0). Another important differences are in Fit Statistics and Type III tests of fixed effects

Fit Statistics (from Dual Quasi-Newton technique)

-2 Res Log Likelihood       24037.76

AIC  (smaller is better)    24053.76

AICC (smaller is better)    24053.80

BIC  (smaller is better)    24082.84

CAIC (smaller is better)    24090.84

HQIC (smaller is better)    24065.42

Generalized Chi-Square      424340.5

Gener. Chi-Square / DF        126.74

Fit Statistics ( Newton-Raphson with Ridging technique)

-2 Res Log Likelihood       28283.79

AIC  (smaller is better)    28299.79

AICC (smaller is better)    28299.83

BIC  (smaller is better)    28328.87

CAIC (smaller is better)    28336.87

HQIC (smaller is better)    28311.45

Generalized Chi-Square      939474.5

Gener. Chi-Square / DF        280.61

Type III Tests of Fixed Effects (from Dual Quasi-Newton technique)

Num      Den

Effect                 DF       DF    F Value    Pr > F

TIME                           1     1048      90.45    <.0001

TIME*TIME                  1     2540       0.84    0.3598

SEX                            1    131.9       4.35    0.0389

AGE                           1    131.9       1.97    0.1626

SEX*AGE                    1    131.9       5.22    0.0240

TIME*SEX                    1     1048       0.03    0.8587

TIME*AGE                   1     1048       0.00    0.9442

TIME*SEX*AGE            1     1048       0.09    0.7703

TIME*TIME*SEX           1     2540       0.10    0.7469

TIME*TIME*AGE           1     2540       0.01    0.9334

TIME*TIME*SEX*AGE       1     2540       0.00    0.9970

Type III Tests of Fixed Effects  ( Newton-Raphson with Ridging technique)

Num      Den

Effect                 DF       DF    F Value    Pr > F

TIME                          1      537      13.28    0.0003

TIME*TIME                  1     3348       0.76    0.3819

SEX                             1    135.9       0.10    0.7577

AGE                             1    135.9       0.05    0.8315

SEX*AGE                      1    135.9       0.12    0.7270

TIME*SEX                      1      537       0.14    0.7074

TIME*AGE                      1      537       0.03    0.8522

TIME*SEX*AGE               1      537       0.03    0.8736

TIME*TIME*SEX               1     3348       0.00    0.9580

TIME*TIME*AGE              1     3348       0.00    0.9760

TIME*TIME*SEX*AGE       1     3348       0.00    0.9762

WHY is such as dramatic difference between these two technique when incorporate covariance structure (i.e. random _residual_ statement) with spatial power structure. (I am not sure same effect will be observed if we use ar(1) or any other covariance strucute, BUT definitely it happened with sp(pow)(xtime) covariance structure) AND what is this criterion difference from  (GCONV=1E-8) to  (XCONV=0).

According to LITTLE 2006,  Newton-Raphson with Ridging technique makes PROC GLIMMIX almost identical to PROC MIXED.

I really feel that i am asking very exhausting questions. Thanks for your patience and all the help.

Regards

Sandhu

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

Without extensive investigation I can't guarantee that this is what is occurring, but I think the NRRIDG technique is finding a saddlepoint in the response surface.  To investigate this, you might save the covariance parameter estimates from the QUANEW and use them as starting values.

But what is very intriguing is the large change in the denominator degrees of freedom.  This is undoubtedly due to the Kenward-Rogers adjustment, and QUANEW resulting in at least one of the G-side covariance parameters being set to zero.  Which parameter has a zero estimate?  Can it be removed, and have the model still make sense?  If it is the intercept term, consider that your R-side model may be exhausting all of the variation, leaving none to be estimated by the intercept.  If this is the case, I think you can safely remove "int" from the first random statement.  If not, could you please post all of the variance parameters from both techniques?

Steve Denham

Occasional Contributor
Posts: 10

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Dear Steve Denham,

I can not figure it out that what is causing such a large change with use of these two techniques. I think you are the best person to understand it. Therefore, i am sending code, model information, dimensions, Optimization Information, LOG etc. Interestingly, When i remove intercept, the NRRIDG technique does not work and following error is reported ERROR: "NRRIDG Optimization cannot be completed.  Optimization routine cannot improve the function value" whereas in same condition i.e. without intercept, QUANEW technique works and it converges.

A) CODE (NRRIDG with random intercept)

proc glimmix data=data2 ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

random _residual_ / sub=id type=sp(pow)(xtime);

nloptions tech= NRRIDG;

run;

Model Information

Data Set                                WORK.DATA2

Response Variable                  VAS

Response Distribution             Gaussian

Variance Function                    Default

Variance Matrix Blocked By     ID

Estimation Technique               Restricted Maximum Likelihood

Degrees of Freedom Method     Kenward-Roger

Fixed Effects SE Adjustment    Kenward-Roger

Dimensions

G-side Cov. Parameters           6

R-side Cov. Parameters           2

Columns in X                          27

Columns in Z per Subject         3

Subjects (Blocks in V)            280

Max Obs per Subject             12

Optimization Information

Optimization Technique         Newton-Raphson with Ridging

Parameters in Optimization    7

Lower Boundaries                  4

Upper Boundaries                  1

Fixed Effects                         Profiled

Residual Variance                  Profiled

Starting From                       Data

Iteration History

Objective                         Max

Iteration    Restarts    Evaluations        Function          Change    Gradient

0           0              4    30662.837656       .                     8506.067

1           0              2    29840.553145    822.28451046    4711.142

2           0              2    29369.643328    470.90981779    3582.741

3           0              2    28916.249747    453.39358027    3044.089

4           0              3    28783.541782    132.70796514    2944.675

5           0              2     28280.91676    502.62502243    8416.694

6           1              5    28751.288165    -470.3714047    7813.525

Convergence criterion (XCONV=0) satisfied.

Covariance Parameter Estimates

Standard

Cov Parm    Subject    Estimate       Error

UN(1,1)     ID          52.0689           .

UN(2,1)     ID           1.0951           .

UN(2,2)     ID          25.7652           .

UN(3,1)     ID          -2.3731           .

UN(3,2)     ID         -10.2573           .

UN(3,3)     ID          12.3989           .

SP(POW) ID           0.5592     0.03979

Residual                 280.30     32.7243

LOG

NOTE: Convergence criterion (XCONV=0) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           10.49 seconds

cpu time            10.43 seconds

B) CODE (NRRIDG without random intercept)

proc glimmix data=data2 ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random time time*time/ subject=id type=un v;

random _residual_ / sub=id type=sp(pow)(xtime);

nloptions tech=  NRRIDG;

run;

Model information remains same

Dimensions

G-side Cov. Parameters           3

R-side Cov. Parameters           2

Columns in X                          27

Columns in Z per Subject         2

Subjects (Blocks in V)            280

Max Obs per Subject             12

Optimization Information

Optimization Technique          Newton-Raphson with Ridging

Parameters in Optimization    4

Lower Boundaries                 3

Upper Boundaries                 1

Fixed Effects                       Profiled

Residual Variance                Profiled

Starting From                      Data

Iteration History

Objective                         Max

Iteration    Restarts    Evaluations        Function          Change    Gradient

0           0              4     29327.32795       .                    11583.9

1           0              2    33917.845247    -4590.517296    11739.39

Optimization routine cannot improve the function value.

Covariance Parameter Estimates

Standard

Cov Parm    Subject    Estimate       Error

UN(1,1)     ID          17.6726           .

UN(2,1)     ID           1.4510           .

UN(2,2)     ID          33.5211           .

SP(POW) ID           0.9000           .

Residual                1415.97           .

LOG

ERROR: NRRIDG Optimization cannot be completed.

NOTE: Optimization routine cannot improve the function value.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           1.43 seconds

cpu time            1.37 seconds

C) CODE (Dual Quasi-Newton with random intercept)

proc glimmix data=data2 ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=un v;

random _residual_ / sub=id type=sp(pow)(xtime);

run;

Model information remains same

Dimensions

G-side Cov. Parameters           6

R-side Cov. Parameters           2

Columns in X                          27

Columns in Z per Subject         3

Subjects (Blocks in V)            280

Max Obs per Subject             12

Optimization Information

Optimization Technique         Dual Quasi-Newton

Parameters in Optimization    7

Lower Boundaries                  4

Upper Boundaries                  1

Fixed Effects                         Profiled

Residual Variance                 Profiled

Starting From                        Data

Iteration History

Objective                         Max

Iteration    Restarts    Evaluations        Function          Change    Gradient

0           0              4    30662.837656       .                    8506.067

1           0              5    28668.856677    1993.9809793    2199.816

2           0             58    26208.493502    2460.3631744    263595.8

3           0              5    26208.485885      0.00761685    270900.4

4           0              3    26188.347277     20.13860824    308956.5

5           0              4    26158.678073     29.66920398      616998

6           0              5    26140.803057     17.87501606    654371.3

7           0              5    26124.120621     16.68243649      688251

8           0              5    26107.310639     16.80998203    722089.4

9           0              5    26090.520122     16.79051661    755560.3

10           0              5    26073.888382     16.63173957    788085.5

11           0              5    26057.541672     16.34671061    818876.5

12           0              5    26041.594314     15.94735789    846943.3

13           0              5    26026.150001     15.44431333    871095.6

14           0              5    26011.302731     14.84726934    889940.6

15           0              2    25964.073338     47.22939308    11515841

16           1             20    25867.099209     96.97412875    635971.6

17           1              3    25828.607906     38.49130370    51872.11

18           1             11    25635.995456    192.61245002     6136433

19           1              7    25632.646006      3.34945002    13163228

20           1              6     25628.79396      3.85204595    13931469

21           1              3     25605.60941     23.18454998    10127958

22           1              3    25568.842337     36.76707290     4627900

.

.

.

.

80           2              5    23953.312146      0.00000207    63380.05

81           2              4    23953.311488      0.00065810    59514.87

82           2              3    23953.311486      0.00000236    53165.53

Convergence criterion (GCONV=1E-8) satisfied.

Estimated G matrix is not positive definite.

Covariance Parameter Estimates

Standard

Cov Parm    Subject    Estimate       Error

UN(1,1)     ID         9.86E-22           .

UN(2,1)     ID          -3.9026      0.8652

UN(2,2)     ID           2.4122      0.2092

UN(3,1)     ID           0.2268     0.06188

UN(3,2)     ID          -0.1050    0.008000

UN(3,3)     ID         7.877E-8           .

SP(POW) ID          0.09958     0.01372

Residual                 140.27      4.8014

LOG

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           25.57 seconds

cpu time            25.30 seconds

D) CODE (Dual Quasi-Newton without random intercept)

proc glimmix data=data2 ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random time time*time/ subject=id type=un v;

random _residual_ / sub=id type=sp(pow)(xtime);

run;

Model information remains same

Dimensions

G-side Cov. Parameters           3

R-side Cov. Parameters           2

Columns in X                          27

Columns in Z per Subject         2

Subjects (Blocks in V)            280

Max Obs per Subject             12

Optimization Information

Optimization Technique          Dual Quasi-Newton

Parameters in Optimization    4

Lower Boundaries                  3

Upper Boundaries                  1

Fixed Effects                         Profiled

Residual Variance                  Profiled

Starting From                        Data

Iteration History

Objective                         Max

Iteration    Restarts    Evaluations        Function          Change    Gradient

0           0              4     29327.32795       .                       11583.9

1           0              5    29281.411327     45.91662314    12868.42

2           0              5    29240.295101     41.11622587    14297.88

3           0             70    26042.998139    3197.2969629    5.6897E9

4           0              4    25979.595376     63.40276297     9.877E8

5           0              3     25957.25277     22.34260529    1.139E10

6           0              3    25939.317553     17.93521689     1.29E10

7           0             18    25841.391645     97.92590789    7.1394E8

8           0             11    25819.989261     21.40238439    9.1343E8

9           0             10    25767.475697     52.51356388    1.5118E9

10           0              6     25765.64274      1.83295751    2.6814E9

11           0              4    25735.316611     30.32612824    3.7442E9

12           0              2    25707.023322     28.29328991    7.3018E9

13           0              2    25665.354323     41.66899876    1.3357E9

14           0              5    25649.042009     16.31231406     5.624E9

15           0              4     25599.39337     49.64863832    1.5771E9

16           0              3    25591.215318      8.17805200    4.4921E9

17           0              4     25575.47631     15.73900814    1.106E10

18           0              2    25550.179798     25.29651182     3.693E9

19           0              5    25522.241508     27.93829080    6.2534E9

20           0              3    25496.791043     25.45046438    6.5937E9

21           0              2    25471.721861     25.06918217    2.1984E9

22           0              9    25439.671707     32.05015405    2.8937E9

.

.

.

130           0              2    24125.065516     22.98137596    1.1381E8

131           0              3    24112.207432     12.85808339    1.2729E8

132           0              5    24107.301953      4.90547927    55729947

133           0              3    24106.054818      1.24713534    50910488

134           0              3    24105.477952      0.57686577     9767172

135           0              3    24105.407721      0.07023051    993324.6

136           0              3     24105.40601      0.00171169    106808.8

137           0              3    24105.405963      0.00004618    89331.43

Convergence criterion (GCONV=1E-8) satisfied.

Estimated G matrix is not positive definite.

LOG

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           29.81 seconds

cpu time            29.12 seconds

I hope my presentation of above facts is clear. Please let me know if i can provide any additional information.

Thanks.

Regards,

Sandhu

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

There is a reason that the default technique for optimization in GLIMMIX is the dual quasi-Newton method.  It is a superior technique, and is finding an optimum solution, while the ridging method is not.  You might be able to force the ridging technique by increasing the ridge parameter.

What is interesting to me is that two of the variances are essentially zero (UN(1,1) and UN(3,3)).  This implies that the random effects are overspecified

proc glimmix data=data2 ORDER=Data ;

class id sex age xtime;

model vas = time time*time

sex age sex*age

sex*time age*time sex*age*time

sex*time*time age*time*time sex*age*time*time / ddfm=kr s cl ;

random int time time*time/ subject=id type=chol v; /* Cholesky parameterization of the unstructured matrix */

random _residual_ / sub=id type=sp(pow)(xtime);

run;

If this still results in zeroes for the id and quadratic term variances, then I think some more thought needs to be given to whether the random _residual_  should be included or not.  I think there is an internal contradiction between the specification of time as a continuous covariate with a random effect and a power law diminution of variation using the spatial power structure.  In other words, we are overthinking the whole process.  Perhaps a simpler approach will tell us all that we need.

Steve Denham

New Contributor
Posts: 3

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Dears Steve and Sandhu,

While I do not have access to your data to make more concrete suggestions, I am not clear why you need to structure the R-side covariance matrix to the extent of what you have "type=sp(pow)(xtime)". Is the outcome variable, pain assessment using 100 mm VAS scale, measured  with a lot of error that could not be accounted for using a diagonal covariance matrix as in "random _residual_/ sub=id type=VC;"? I would go with a well structured G-side covariance matrix with the default R-side (you still have a non-diagonal V matrix that accounts for both with-in and between group covariation, BTW) for both model parsimony and computational simplicity purposes. Thanks.

Mulugeta

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

Thanks for pulling us back from the brink.  This is where the last two lines that I wrote was leading, but this is much more to the point, and WAAAAAY more likely to give interpretable results.

Steve Denham

Occasional Contributor
Posts: 10

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Dear Steve and Mulugeta,

I tried both the alternatives as recommended by both of you.  By using Cholesky parameterization of the unstructured matrix, the only the quadratic term variances is zero, and not the id. But the use of random _residual_/ sub=id type=VC did not improve the model and it still resulted in zeroes for the id and quadratic term variances.

A) using Steve's code with Cholesky parameterization of the unstructured matrix

Covariance Parameter Estimates

Standard

Cov Parm     Subject    Estimate       Error

CHOL(1,1)    ID           3.1811      0.4606

CHOL(2,1)    ID          -0.1228      0.1851

CHOL(2,2)    ID         0.000026      0.1909

CHOL(3,1)    ID         -0.00994     0.01326

CHOL(3,2)    ID         -2.09E-6     0.01529

CHOL(3,3)    ID         4.38E-18           .

SP(POW)      ID          -0.1072     0.01474

Residual                  115.98      3.1069

Fit Statistics

-2 Res Log Likelihood       24954.65

AIC  (smaller is better)    24968.65

AICC (smaller is better)    24968.68

BIC  (smaller is better)    24994.09

CAIC (smaller is better)    25001.09

HQIC (smaller is better)    24978.85

Generalized Chi-Square      388284.9

Gener. Chi-Square / DF        115.98

it converged in 49 Iteration

Log:

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           25.71 seconds

cpu time            25.36 seconds

B) using Mulugeta's code with diagonal covariance matrix as in "random _residual_/ sub=id type=VC and unstructured matrix UN variance for random effect

Covariance Parameter Estimates

Standard

Cov Parm         Subject    Estimate       Error

UN(1,1)          ID         3.23E-18           .

UN(2,1)          ID           0.1491      0.9892

UN(2,2)          ID           1.5762      0.1915

UN(3,1)          ID         -0.04960     0.07431

UN(3,2)          ID         -0.07712    0.007403

UN(3,3)          ID         0.000025           .

Residual (VC)                 149.52      3.9873

Fit Statistics

-2 Res Log Likelihood       25206.89

AIC  (smaller is better)    25218.89

AICC (smaller is better)    25218.92

BIC  (smaller is better)    25240.70

CAIC (smaller is better)    25246.70

HQIC (smaller is better)    25227.64

it converged in 53 iterations

LOG

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           17.31 seconds

cpu time            17.05 seconds

B) using Mulugeta's code with  diagonal covariance matrix as in "random _residual_/ sub=id type=VC and Cholesky parameterization of the unstructured matrix variance for random effect

Covariance Parameter Estimates

Standard

Cov Parm         Subject    Estimate       Error

CHOL(1,1)        ID           0.6879      0.8774

CHOL(2,1)        ID           0.1156      0.2015

CHOL(2,2)        ID          4.1E-19           .

CHOL(3,1)        ID         -0.01065     0.01730

CHOL(3,2)        ID         0.000451    0.003211

CHOL(3,3)        ID         2.13E-16           .

Residual (VC)                 139.60      3.6281

Fit Statistics

-2 Res Log Likelihood       26165.98

AIC  (smaller is better)    26175.98

AICC (smaller is better)    26176.00

BIC  (smaller is better)    26194.15

CAIC (smaller is better)    26199.15

HQIC (smaller is better)    26183.27

Generalized Chi-Square      467391.5

Gener. Chi-Square / DF        139.60

it converged in 41 iterations

Log

NOTE: Convergence criterion (GCONV=1E-8) satisfied.

NOTE: At least one element of the gradient is greater than 1e-3.

NOTE: Estimated G matrix is not positive definite.

NOTE: PROCEDURE GLIMMIX used (Total process time):

real time           3.25 seconds

cpu time            3.19 seconds

Thanks

Regards

Sandhu

Posts: 2,655

## Re: Nonlinear Mixed Effect Model Analysis

It is no great crime if a variance component is zero (G matrix is not positive definite).  It simply means that for the data available, other terms explain all of the variation.  So long as the proper adjustment to degrees of freedom is applied, tests and confidence intervals should be acceptable.

Given all that has gone on here, it appears that the Cholesky parameterization leads to the model with the smallest information criteria, so I would just go with that.  It implies that after the individual variability, linear in time variability, their covariance, and their covariance with quadratic in time are accounted for, you have no remaining variability due strictly to the quadratic in time component.  So long as you feel that the covariances are important, then this is where you end up.  With a larger or different data set, you might get something estimated for the quadratic component.

If you believe, a priori, that the covariances are not important, then you should steer away from the unstructured covariance types, and use the diagonal structure that Muller proposed.

Steve Denham

Occasional Contributor
Posts: 10

## Re: Nonlinear Mixed Effect Model Analysis

Posted in reply to SteveDenham

Thanks a lot, Steve. I will do that.