Hi, I'm currently using the scale() function in a glmer model (see below) and I'm having some inconsistencies between my standardized coefficient estimates from SAS and those in R. In SAS, the STDCOEF option in a generalized linear mixed model reports solutions for fixed effects in terms of the standardized (scaled and/or centered) coefficients. See SAS code below. Is there a reason why the standardized coefficient estimates are different between the R and SAS output? I assumed the standardization process was the same. Thanks for the help!

R Code:


results<-glmer(R0A1~scale(Dist_MP)+scale(Dist_MPHW)+scale(Dist_HW)+scale(Dist_YP)+scale(Dist_AG)+scale(Dist_Shrub)+scale(Dist_CountyRoad)+scale(Dist_PrimaryRoad)+scale(Dist_SecondaryRoad)+scale(Dist_TertFireRoad)+(1|ID),data=secondorder,family=binomial)

Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: binomial ( logit )
Formula:
R0A1 ~ scale(Dist_MP) + scale(Dist_MPHW) + scale(Dist_HW) + scale(Dist_YP) +
scale(Dist_AG) + scale(Dist_Shrub) + scale(Dist_CountyRoad) +
scale(Dist_PrimaryRoad) + scale(Dist_SecondaryRoad) + scale(Dist_TertFireRoad) +
(1 | ID)
Data: secondorder

AIC BIC logLik deviance df.resid
48038.2 48140.2 -24007.1 48014.2 36374

Scaled residuals:
Min 1Q Median 3Q Max
-1.7613 -0.9819 0.2380 0.9124 3.5661

Random effects:
Groups Name Variance Std.Dev.
ID (Intercept) 0.008305 0.09113
Number of obs: 36386, groups: ID, 49

Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.003954 0.018385 -0.215 0.829699
scale(Dist_MP) -0.172902 0.013084 -13.215 < 2e-16 ***
scale(Dist_MPHW) -0.214762 0.011945 -17.980 < 2e-16 ***
scale(Dist_HW) -0.119081 0.011116 -10.712 < 2e-16 ***
scale(Dist_YP) 0.068134 0.013078 5.210 1.89e-07 ***
scale(Dist_AG) -0.196623 0.011886 -16.542 < 2e-16 ***
scale(Dist_Shrub) -0.103676 0.011642 -8.905 < 2e-16 ***
scale(Dist_CountyRoad) -0.078235 0.011956 -6.544 6.01e-11 ***
scale(Dist_PrimaryRoad) -0.259842 0.011994 -21.664 < 2e-16 ***
scale(Dist_SecondaryRoad) -0.136900 0.012455 -10.992 < 2e-16 ***
scale(Dist_TertFireRoad) -0.047486 0.012365 -3.840 0.000123 ***
---
Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1

Correlation of Fixed Effects:
(Intr) s(D_MP) s(D_MPH s(D_HW s(D_YP s(D_AG s(D_S) s(D_CR s(D_PR
scl(Dst_MP) 0.031
scl(D_MPHW) 0.026 0.005
scl(Dst_HW) 0.006 0.147 0.072
scl(Dst_YP) 0.040 0.215 0.069 0.001
scl(Dst_AG) 0.001 0.089 0.035 0.045 -0.108
scl(Dst_Sh) -0.005 0.007 -0.031 0.020 0.046 -0.178
scl(Dst_CR) 0.016 0.032 -0.006 -0.036 0.111 0.036 -0.025
scl(Dst_PR) 0.005 -0.111 -0.069 0.000 -0.089 -0.022 -0.096 0.010
scl(Dst_SR) -0.009 -0.112 -0.034 0.045 0.099 -0.064 -0.073 -0.085 -0.036
scl(Ds_TFR) -0.014 -0.126 -0.044 -0.064 -0.360 -0.133 -0.032 -0.050 -0.007
s(D_SR
scl(Dst_MP)
scl(D_MPHW)
scl(Dst_HW)
scl(Dst_YP)
scl(Dst_AG)
scl(Dst_Sh)
scl(Dst_CR)
scl(Dst_PR)
scl(Dst_SR)
scl(Ds_TFR) 0.144

SAS Code:

PROC GLIMMIX DATA=PROJECT;
CLASS ID;
MODEL R0A1 (EVENT='1') = DIST_MP DIST_MPHW DIST_HW DIST_YP DIST_AG DIST_SHRUB DIST_COUNTYROAD DIST_PRIMARYROAD DIST_SECONDARYROAD DIST_TERTFIREROAD/SOLUTION DIST=BINARY LINK=LOGIT STDCOEF;
RANDOM INTERCEPT / SUBJECT=ID TYPE=VC;
NLOPTIONS GCONV=0;
RUN;


Name:  sasOUTPUT.jpg
Views: 47
Size:  16.7 KB