[R-sig-ME] Reproducing results from an old lmer fit

Afshartous, David DAfshartous at med.miami.edu
Thu Feb 26 17:31:08 CET 2009


For a paper revision I'm trying to reproduce some results from an old lmer
fit with Rv2.7.1 prior to 5/28/08.  However, when I currently load Rv2.7.1
and lmer, the variance component estimates are slightly different than the
original fit; the sessionInfo() is as follows:

> sessionInfo()
R version 2.7.1 (2008-06-23)


attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] lme4_0.999375-24   Matrix_0.999375-11 lattice_0.17-8

loaded via a namespace (and not attached):
[1] grid_2.7.1  nlme_3.1-89

Thus, I assume that I need to use the same older version of lme4 and/or
Matrix which might be responsible for the difference in the results?  If
this is possible, how is this done?


PS - for whatever it's worth, if I do the fit with lme (nlme_3.1-89) under
Rv2.7.1 the results are closer to the original lmer results.


Original lmer fit from 5/08:
Model 2:
AIC  BIC logLik MLdeviance REMLdeviance
 2813 2843  -1397       2829         2795
Random effects:
 Groups     Name            Variance Std.Dev. Corr
 subject   (Intercept)      2226.3   47.183
            Drug            2132.9   46.184  -0.865
 Residual                   13673.6  116.934

Current lmer fit:
 AIC  BIC logLik deviance REMLdev
 2815 2849  -1397     2830    2795
Random effects:
 Groups     Name               Variance Std.Dev. Corr
 Patient_no (Intercept)         2165.1   46.531
            Drug.full.reverseC  1386.3   37.233  -1.000
 Residual                      13947.5  118.100

Current lme fit:
   AIC      BIC    logLik
  2814.611 2848.638 -1397.305

                   StdDev    Corr
(Intercept)         47.21031 (Intr)
Drug.full.reverseC  46.14014 -0.866
Residual           116.93541

More information about the R-sig-mixed-models mailing list