You can use proc GLM which considering missing values and gave you mean, std div, std error, cv, min, max, ......

I think you just need to do it manually. You can use an LS means statement to get a mean and SE for each treatment level of interest.

You can do it manually, using the square root of the RESIDUAL from the Covariance Parameter  Estimates table and a mean of the LSMEANS estimates.

I have an automatized solution Moshood Bakare ...Check my SAS routine down.

/*Obtaining Coeficiente of variation in SAS - PROC MIXED*/

/*By: Alysson Jalles / [email protected]*/

/*TIP:

Code in Uppercase you can change

Code in lowercase you cannot change*/

data DATASET_RCBD;

input TREAT REP VAR1;

cards;

1 1 10.5

1 2 12.5

2 1 9

2 2 8.5

;

Title "PROC GLM - Coeficient of Variation (%)";

proc glm data=DATASET_RCBD;

class TREAT REP;

model VAR1=TREAT REP/ss3;

ods select fitstatistics;

ods output fitstatistics=PROC_GLM_CV (keep=CV);

run;

data PROC_GLM_CV; set PROC_GLM_CV;

label CV='Coeficient of Variation (%)';

run;

Title "PROC MIXED - Coeficient of Variation (%)";

proc mixed data=DATASET_RCBD;

class TREAT REP;

model VAR1=REP;

random TREAT;

estimate "General Mean with confidence interval" intercept 1| TREAT 0/cl;

ods select covparms;

*Exporting RESIDUAL COVARIANCE and General Mean;

ods output covparms=RESIDUAL_COV (rename=(estimate=RESID))

estimates=OVERALL_MEAN (keep=estimate rename=(estimate=MEAN));

run;

/*Merging covariance dataset with mean dataset*/

data RESIDUAL_COV;set RESIDUAL_COV;

if CovParm='TREAT' then delete; run;

data PROC_MIXED_CV; merge RESIDUAL_COV OVERALL_MEAN;

CV=sqrt(RESID)*100/MEAN;

label CV='Coeficient of Variation (%)';

drop COVPARM RESID MEAN;

run;

/******PRINTING RESULTS OF CV*********************/

ods preferences;

Title "Proc GLM CV"; run;

Proc print data=PROC_GLM_CV noobs label; run;

Title "Proc MIXED CV"; run;

Proc print data=PROC_MIXED_CV noobs label; run;

*Moshood Bakare  - If possible vote to promote my answer;

Hi,

I have a linear mixed models (LMM) to analyse my RCBD.

is there a way to derive a coefficient of variance (CoV) in R for LMM? I have the mean estimates for each treatment (fixed effect) but I also need to report the CoV for each treatments (fixed effect). Is there a way of doing so?