Skip to content
Prev 6901 / 20628 Next

Zhang 2011 (re)analysis

Here is some comparison between glm,  glmer (using lme4 with Laplace)
and sabreR (which uses AGHQ, if I recall correctly).
The glm analysis replicates exactly the results reported in Everitt
and Hothorn (2002, Table 13.1, around page 175).  Thus, I am pretty
sure I am using the correct data.

 sabreR gives estimates both for the standard homogeneous model,
replicating the glm(),  as well as a random effects model, replicating
glmer() pretty closely, I think

Reinhold
Call:
glm(formula = status.b ~ centr.b + treat.b + sex.b + pre.b +
    age, family = "binomial", data = resp)

Deviance Residuals:
    Min       1Q   Median       3Q      Max
-2.3146  -0.8551   0.4336   0.8953   1.9246

Coefficients:
             Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.900171   0.337653  -2.666  0.00768 **
centr.b      0.671601   0.239567   2.803  0.00506 **
treat.b      1.299216   0.236841   5.486 4.12e-08 ***
sex.b        0.119244   0.294671   0.405  0.68572
pre.b        1.882029   0.241290   7.800 6.20e-15 ***
age         -0.018166   0.008864  -2.049  0.04043 *
---
Signif. codes:  0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 608.93  on 443  degrees of freedom
Residual deviance: 483.22  on 438  degrees of freedom
AIC: 495.22

Number of Fisher Scoring iterations: 4
+       family=binomial,data=resp), cor=FALSE)
Generalized linear mixed model fit by the Laplace approximation
Formula: status ~ centre + treatment + sex + baseline + age + (1 | subject)
   Data: resp
 AIC   BIC logLik deviance
 443 471.7 -214.5      429
Random effects:
 Groups  Name        Variance Std.Dev.
 subject (Intercept) 3.8647   1.9659
Number of obs: 444, groups: subject, 111

Fixed effects:
                   Estimate Std. Error z value Pr(>|z|)
(Intercept)        -1.64438    0.75829  -2.169   0.0301 *
centre2             1.04382    0.53193   1.962   0.0497 *
treatmenttreatment  2.15746    0.51757   4.168 3.07e-05 ***
sexmale             0.20194    0.66117   0.305   0.7600
baselinegood        3.06990    0.52608   5.835 5.37e-09 ***
age                -0.02540    0.01998  -1.271   0.2037
---
Signif. codes:  0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
# ... deleted some output
(Standard Homogenous Model)
    Parameter              Estimate         Std. Err.        Z-score
    ____________________________________________________________________
    (intercept)           -0.90017          0.33765          -2.6660
    centr.b                0.67160          0.23957           2.8034
    treat.b                 1.2992          0.23684           5.4856
    sex.b                  0.11924          0.29467          0.40467
    pre.b                   1.8820          0.24129           7.7999
    age                   -0.18166E-01      0.88644E-02      -2.0493



(Random Effects Model)
    Parameter              Estimate         Std. Err.        Z-score
    ____________________________________________________________________
    (intercept)            -1.6642          0.84652          -1.9660
    centr.b                0.99044          0.56561           1.7511
    treat.b                 2.1265          0.57198           3.7177
    sex.b                  0.18166          0.70814          0.25653
    pre.b                   2.9987          0.60174           4.9834
    age                   -0.22949E-01      0.21337E-01      -1.0755
    scale                   1.9955          0.32093           6.2180

# ... deleted some output

        
On Mon, Oct 31, 2011 at 2:10 AM, Ben Bolker <bbolker at gmail.com> wrote: