Proc Glimmix ERROR: LINES display is not produced for the PD effect because adjusted p-values are no

[zelapd20]

Hi all,

Running a statistical analysis of my data for a three way interaction in proc glimmix and ran into an error 

"ERROR: LINES display is not produced for the PD effect because adjusted p-values are not
       available for all least-squares means differences."

The error popped up for the three factors and all the two way interactions but not for the three way interaction. Since the three way interaction is non-significant, it doesn't help for my interpretation.

Any help is appreciated.

ERROR: LINES display is not produced for the PD effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the F effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because adjusted p-values are not
       available for all least-squares means differences.

[PaigeMiller]

You need to show us the entire LOG for PROC GLIMMIX, not just the error messages. Please help us by maintaining the formatting of the log making it easier to read; copy the entire PROC GLIMMIX log as text and paste it into the window that appears when you click on the {i} icon. DO NOT SKIP THIS STEP.

Also please show us the output that you do get from PROC GLIMMIX. Copy as an image and paste is fine here.

[zelapd20]

685  DATA CAPDVFF2019;
686  INPUT  PD$ V$ F$ Rep DSL AUDPC rAUDPC;
687  cards;

NOTE: SAS went to a new line when INPUT statement reached past the end of a line.
NOTE: The data set WORK.CAPDVFF2019 has 144 observations and 7 variables.
NOTE: DATA statement used (Total process time):
      real time           0.02 seconds
      cpu time            0.01 seconds


833  run;
NOTE: Writing HTML Body file: sashtml1.htm
834  PROC glimmix  DATA=CAPDVFF2019 plots=pearsonpanel;
835  CLASS PD V F;
836  MODEL  DSL =  PD V F PD*V PD*F V*F PD*V*F/DDFM=KENWARDROGER;
837  RANDOM Rep ;
838  lsmeans PD V F PD*V PD*F V*F PD*V*F/pdiff adjust=tukey lines;
839  run;

NOTE: With DDFM=SATTERTHWAITE or DDFM=KENWARDROGER, unadjusted p-values in tests are based on the
      degrees of freedom specific to that comparison. P-values that are adjusted for
      multiplicity, however, are by default based on the denominator degrees of freedom for the
      Type III test of the fixed effect. If you specify the ADJDFE=ROW option in the LSMEANS or
      LSMESTIMATE statement, the adjusted p-values take into account the row-wise degrees of
      freedom.


NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.
ERROR: LINES display is not produced for the PD effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the F effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because adjusted p-values are not
       available for all least-squares means differences.
NOTE: PROCEDURE GLIMMIX used (Total process time):
      real time           3.84 seconds
      cpu time            2.71 seconds


840  PROC glimmix  DATA=CAPDVFF2019 plots=pearsonpanel;
841  CLASS PD V F;
842  MODEL  AUDPC =  PD V F PD*V PD*F V*F PD*V*F/DDFM=KENWARDROGER;
843  RANDOM Rep ;
844  lsmeans PD V F PD*V PD*F V*F PD*V*F/pdiff adjust=tukey lines;
845  run;

NOTE: With DDFM=SATTERTHWAITE or DDFM=KENWARDROGER, unadjusted p-values in tests are based on the
      degrees of freedom specific to that comparison. P-values that are adjusted for
      multiplicity, however, are by default based on the denominator degrees of freedom for the
      Type III test of the fixed effect. If you specify the ADJDFE=ROW option in the LSMEANS or
      LSMESTIMATE statement, the adjusted p-values take into account the row-wise degrees of
      freedom.


NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.
ERROR: LINES display is not produced for the PD effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the F effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because adjusted p-values are not
       available for all least-squares means differences.
NOTE: PROCEDURE GLIMMIX used (Total process time):
      real time           2.48 seconds
      cpu time            2.28 seconds


846  PROC glimmix  DATA=CAPDVFF2019 plots=pearsonpanel;
847  CLASS PD V F;
848  MODEL  rAUDPC =  PD V F PD*V PD*F V*F PD*V*F/DDFM=KENWARDROGER;
849  RANDOM Rep;
850  lsmeans PD V F PD*V PD*F V*F PD*V*F/pdiff adjust=tukey lines;
851  run;

NOTE: With DDFM=SATTERTHWAITE or DDFM=KENWARDROGER, unadjusted p-values in tests are based on the
      degrees of freedom specific to that comparison. P-values that are adjusted for
      multiplicity, however, are by default based on the denominator degrees of freedom for the
      Type III test of the fixed effect. If you specify the ADJDFE=ROW option in the LSMEANS or
      LSMESTIMATE statement, the adjusted p-values take into account the row-wise degrees of
      freedom.


NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.
ERROR: LINES display is not produced for the PD effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the F effect because adjusted p-values are not available
       for all least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because adjusted p-values are not
       available for all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because adjusted p-values are not
       available for all least-squares means differences.
NOTE: PROCEDURE GLIMMIX used (Total process time):
      real time           2.35 seconds
      cpu time            2.17 seconds


852
853  ods graphics off;
854  ods rtf close;

Thank you.

I put the Log PROC GLIMMIX as requested. The output window on the other hand doesn't show anything. Only results are showing. 

[PaigeMiller]

As an experiment, take out the ADJUST=TUKEY and see if the error goes away.

[zelapd20]

It didn't change. Below is the what is on log error message. If not Tukey, may I need something else for the mean comparison though?

Thank you once again.

855  DATA CAPDVFF2019;
856  INPUT  PD$ V$ F$ Rep DSL AUDPC rAUDPC;
857  cards;

NOTE: SAS went to a new line when INPUT statement reached past the end of a line.
NOTE: The data set WORK.CAPDVFF2019 has 144 observations and 7 variables.
NOTE: DATA statement used (Total process time):
      real time           0.01 seconds
      cpu time            0.00 seconds


1003  run;
NOTE: Writing HTML Body file: sashtml2.htm
1004  PROC glimmix  DATA=CAPDVFF2019 plots=pearsonpanel;
1005  CLASS PD V F;
1006  MODEL  DSL =  PD V F PD*V PD*F V*F PD*V*F/DDFM=KENWARDROGER;
1007  RANDOM Rep ;
1008  lsmeans PD V F PD*V PD*F V*F PD*V*F/pdiff lines;
1009  run;



NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.
ERROR: LINES display is not produced for the PD effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the V effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the F effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because p-values are not available for
       all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because p-values are not available for
       all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because p-values are not available for
       all least-squares means differences.
NOTE: PROCEDURE GLIMMIX used (Total process time):
      real time           2.13 seconds
      cpu time            1.48 seconds


1010  PROC glimmix  DATA=CAPDVFF2019 plots=pearsonpanel;
1011  CLASS PD V F;
1012  MODEL  AUDPC =  PD V F PD*V PD*F V*F PD*V*F/DDFM=KENWARDROGER;
1013  RANDOM Rep ;
1014  lsmeans PD V F PD*V PD*F V*F PD*V*F/pdiff lines;
1015  run;



NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.
ERROR: LINES display is not produced for the PD effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the V effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the F effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because p-values are not available for
       all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because p-values are not available for
       all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because p-values are not available for
       all least-squares means differences.
NOTE: PROCEDURE GLIMMIX used (Total process time):
      real time           0.88 seconds
      cpu time            0.72 seconds


1016  PROC glimmix  DATA=CAPDVFF2019 plots=pearsonpanel;
1017  CLASS PD V F;
1018  MODEL  rAUDPC =  PD V F PD*V PD*F V*F PD*V*F/DDFM=KENWARDROGER;
1019  RANDOM Rep;
1020  lsmeans PD V F PD*V PD*F V*F PD*V*F/pdiff lines;
1021  run;



NOTE: Convergence criterion (ABSGCONV=0.00001) satisfied.
ERROR: LINES display is not produced for the PD effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the V effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the F effect because p-values are not available for all
       least-squares means differences.
ERROR: LINES display is not produced for the PD*V effect because p-values are not available for
       all least-squares means differences.
ERROR: LINES display is not produced for the PD*F effect because p-values are not available for
       all least-squares means differences.
ERROR: LINES display is not produced for the V*F effect because p-values are not available for
       all least-squares means differences.
NOTE: PROCEDURE GLIMMIX used (Total process time):
      real time           0.86 seconds
      cpu time            0.68 seconds


1022
1023  ods graphics off;
1024  ods rtf close;

[PaigeMiller]

How many levels of each of the class variables PD, V and F? How many levels of REP?

[zelapd20]

PD = 6

V = 4

F = 2

Rep = 3

[PaigeMiller]

You have no degrees of freedom for error, so p-values cannot be computed.

Try taking the 3-way interaction out of the model if you want p-values and lines.

[zelapd20]

Thank you. Let me try that. How was the DF calculated 🙂 in the meantime

[zelapd20]

Shall I possibly try a different procedure than GLIMMIX

[PaigeMiller]

It's not the PROC that is the problem, it is your model which has no degrees of freedom for error.

[zelapd20]

It just came back with the same error message

[PaigeMiller]

Do you get any output at all? If not, try running it without LSMEANS and show us the output.

[ashleyzyy]

Hi there,

I had the similar problem, when I removed the ddfm=kr from the model line it gets all solved. but I don't understand why it would like that, does it affect the bias?

here is my origional code. 

thanks, 

data proximate;

input location$ rep GE

;

response=GE;var='GE';output;
cards;

Kansas 1 4255.79
Kansas 2 4338.34
Kansas 3 4268.54
Kansas 4 4239.63
Kansas 5 4201.55
Kansas 6 4230.07
LA 1 3986.36
LA 2 4274.80
LA 3 4024.32
LA 4 4239.64
LA 5 4165.36
LA 6 4243.31
Dallas 1 4239.02
Dallas 2 4195.14
Dallas 3 4151.11
Dallas 4 4097.20
Dallas 5 4127.85
Dallas 6 4177.10
GA 1 4344.53
GA 2 4269.53
GA 3 4336.44
GA 4 4514.79
GA 5 4388.69
GA 6 4237.09
Stockton 1 4099.26
Stockton 2 4010.61
Stockton 3 3960.10
Stockton 4 4024.82
Stockton 5 4238.10
Stockton 6 4104.96
Baltimore 1 4086.10
Baltimore 2 4168.86
Baltimore 3 4064.65
Baltimore 4 4460.57
Baltimore 5 4151.72
Baltimore 6 3982.83
;
RUN;
proc print data=proximate;

run;

proc sort data=proximate;

by var;

run;

proc glimmix data=proximate plots=RESIDUALPANEL;

by var;

class location rep;

model response = location/ddfm=kr;

random rep(location);

lsmeans location / pdiff lines adjust=tukey;

run;

[StatsMan]

You have one observation for each rep and location combination, so the rep(location) effect is confounded with the residual variance. You would need multiple observations for each rep and location combination to estimate a rep(location) effect. Remove the RANDOM statement and you get the expected results for this data and model - including the LINES output