Skip to content

Converting SAS code into R

2 messages · Nicholas Mitsakakis, Wolfgang Viechtbauer

Original
Nicholas Mitsakakis
# 
I have been trying to convert some PROC MIXED SAS code into R, but without
success. The code is:

proc mixed data=rmanova4;class randomization_arm cancer_type site wk;
model chgpf=randomization_arm cancer_type site wk;
repeated  / subject=study_id;
contrast '12 vs 4' randomization_arm 1 -1;
lsmeans randomization_arm / cl pdiff alpha=0.05;
run;quit;

I have tried something like

lme(chgpf ~ Randomization_Arm  + Cancer_Type + site + wk, data=rmanova.data,
            random = ~ 1  | Study_ID,
            correlation = corSymm(form = ~ 1 | Study_ID),
            na.action=na.exclude, method = "REML")

but I am getting different estimate values.

Perhaps I am misunderstanding something basic. Any comment/suggestion would
be greatly appreciated.

I am adding here the output. Part of the output from the SAS code is below:

Least Squares Means
Effect  Randomization_Arm   Estimate    Standard Error  DF  t Value Pr
Randomization_Arm   12 weekly BTA   -4.5441 1.3163  222 -3.45   0.0007
 0.05    -7.1382 -1.9501
Randomization_Arm   4 weekly BTA    -6.4224 1.3143  222 -4.89   <.0001
 0.05    -9.0126 -3.8322

Differences of Least Squares Means
Effect  Randomization_Arm   _Randomization_Arm  Estimate    Standard
Error  DF  t Value Pr > |t|    Alpha   Lower   Upper
Randomization_Arm   12 weekly BTA   4 weekly BTA    1.8783  1.4774
222 1.27    0.2049  0.05    -1.0332 4.7898

The output from the R code is below:

Linear mixed-effects model fit by REML
  Data: rmanova.data
       AIC     BIC    logLik
  6526.315 6586.65 -3250.157

Random effects:
 Formula: ~1 | Study_ID
        (Intercept) Residual
StdDev:    16.51816 12.95417

Correlation Structure: General
 Formula: ~1 | Study_ID
 Parameter estimate(s):
 Correlation:
  1      2      3     2  0.222              3 -0.159  0.225       4
-0.421 -0.042  0.083
Fixed effects:  chgpf ~ Randomization_Arm + Cancer_Type + site + wk
                                  Value Std.Error  DF    t-value
p-value(Intercept)                    5.739240 2.8987216 541
1.9799209  0.0482
Randomization_Arm4 weekly BTA -1.174704 2.3915873 225 -0.4911817  0.6238
Cancer_TypeProsta             -4.459715 2.4711025 225 -1.8047469  0.0725
site                          -1.917902 0.9709655 225 -1.9752530  0.0495
wk                            -1.570707 0.5115174 541 -3.0706809  0.0022
 Correlation:
                              (Intr) R_A4wB Cnc_TP site
Randomization_Arm4 weekly BTA -0.440
Cancer_TypeProsta             -0.314  0.043
site                          -0.598  0.003 -0.064
wk                            -0.421 -0.004  0.032  0.003

Standardized Within-Group Residuals:
        Min          Q1         Med          Q3         Max
-4.99530346 -0.36507852  0.08308708  0.45937864  3.12730244

Number of Observations: 771
Number of Groups: 229








--
Nicholas Mitsakakis, MSc, PhD, P.Stat.

Senior Biostatistician and Associate Scientist
Children's Hospital of Eastern Ontario Research Institute
Adjunct Lecturer, Dalla Lana School of Public Health, University of Toronto
Wolfgang Viechtbauer
# 
Dear Nicholas,

The model specified via 'proc mixed' is actually not a mixed-effects / multilevel model at all. The default 'type' for 'repeated' is VC, which is the same as in a standard regression type model (i.e., it assumes independence between repeated observations within subjects and assumes a single variance error variance component). So you are in essence just fitting:

proc glm data=rmanova4; class randomization_arm cancer_type site wk;
 model chgpf = randomization_arm cancer_type site wk;
run;

which is identical in R to:

res <- lm(chgpf ~ Randomization_Arm + Cancer_Type + site + wk, data=rmanova.data)
summary(res)

(assuming that the the predictors are all coded as factors in R).

I suspect that you actually wanted to fit a different model in SAS to begin with (why else use proc mixed?), but you would have to tell us first what the actual model is that you wanted to fit in SAS before we can tell you how to fit the same model in R.

Best,
Wolfgang