NEWS
VetResearchLMM 2.0.0
Changes visible to users
- The data sets
ex121, ex124, ex125, ex127, ex31, ex32 and
ex33, the help topics Examp1.3.2 to Examp3.3, and
emmeans_mixed_model() and report_mixed_model() keep their names,
columns and arguments.
emmeans_mixed_model() and report_mixed_model() are S3 generics:
their .default methods behave as in version 1.1.0 (models of
lme4 and nlme, through emmeans and report), and new
methods accept results of vlm_fit().
- The help topics
Examp* now show the SAS program of the book next to
the same analysis with the functions of the package, which reproduces
the numbers the book prints; lme4 remains as an optional
comparison.
- The four vignettes of version 1.1.0 are replaced by six (getting
started, theory, the examples of Chapters 1-2 and of Chapter 3,
validation, application), each also as PDF.
Corrections
ex33: the 11th measurement of BO322 is on day 25 (not a second day
23), that of ND66 on day 25 (not 26), and the identifier "BO326 "
lost its trailing space, as in Appendix A of the book. With the
corrected data the variance components of Example 3.3 are the printed
5.478 and 4.965 (5.522 and 4.943 before), and the first-order
autoregressive model gives the printed rho = 0.621.
New
vlm_fit(): an engine for linear mixed models that gives the results
of SAS PROC MIXED: REML, ML or Type 3 (moment) estimation by
average-information Newton steps with bounds; random variance
components, random coefficients (vlm_random(), VC or UN) and
residual structures (vlm_repeated(): CS, AR(1), SP(POW), UN); the
solution for the fixed effects in the overparameterised coding; Type 3
and Type 1 tests; containment, residual, between-within, Satterthwaite
and Kenward-Roger degrees of freedom; standard errors of the
covariance parameters; AIC, AICC, BIC, the SAS 6 criteria of the book
and the null-model likelihood ratio test.
vlm_lsmeans(): least-squares means at chosen covariate values, with
Tukey-Kramer, Bonferroni and Sidak comparisons.
vlm_estimate() and vlm_contrast(): ESTIMATE and CONTRAST
statements with coefficients given per effect as in SAS, estimability
checks, and the narrow inference space through random-effect
coefficients.
vlm_blup(): best linear unbiased predictions with standard errors of
prediction.
vlm_ems(): the analysis of variance of PROC GLM with a RANDOM
statement: expected mean squares, synthesised error terms with
Satterthwaite degrees of freedom and Type 3 estimates of the variance
components.
vlm_compare(): information criteria and likelihood ratio tests of
several fits.
vlm_interpret(), vlm_cite(), vlm_notation(), vlm_reference(),
plot() and autoplot() methods, and vlm_app(), a shiny
application in which every example of the book loads with one click.
- Data sets
ex122 (Example 1.2.2), ex123 (Example 1.2.3) and ex126
(Example 1.2.6) from Appendix A, and the help topic Examp2.3.2 for
the models of Section 2.3.2.
- 116 numbers printed in the book are recomputed by the tests
(
inst/extdata/reference.csv); 115 agree and the other is printed
truncated. The engine is also checked against lme4, lmerTest,
pbkrtest, nlme and emmeans.
Package
- Imports data.table, ggplot2 and scales; the packages used
for comparisons are suggested.
VetResearchLMM 1.1.0 (2026-05-05)
- Package-level documentation,
report_mixed_model() and
emmeans_mixed_model(), Quarto vignettes and tests of the book
examples.