rmedsem 1.1.0
Breaking changes
summary() returns a summary.rmedsem
objectas.data.frame() returns a plain data framermedsem() stops with an error for models that
previously ran but gave wrong or empty resultsci.two.tailed now determines all intervals- blavaan: the direct effect’s
pval is now the posterior
probability of the opposite sign
mcreps defaults to 5000 and is no longer raised to the
sample size- RID is not reported if the direct effect is not significant;
RIT() warns if the total effect is too small
New features
summary() method- New methods
coef(), confint() and
nobs() for rmedsem objects.
rmedsem.blavaan() gains ci.two.tailed and
hdi (highest-density intervals)- Printing is now an extensible S3 method system
- Result objects gain the elements
zlc.method
- All exported functions check their arguments and stop with
informative messages
- lavaan: a warning is given if the model did not converge.
- cSEM: new
seed argument; results are reproducible with
set.seed()
Bug fixes
- lavaan and modsem models with labelled regression paths (e.g.,
read ~ a*math) now work
- blavaan: the rows of the printed table were shifted relative to
their labels
- modsem: the moderation effects are printed in causal direction
- cSEM: the bootstrap interval of the indirect effect was taken from
the incorrect row
digits and indent are passed correctly to
the effect-size output.plot_coef() no longer emits a
position_dodge() warning- Very small p-values are no longer printed as
0
plot_effect() title and caption are wrapped to fit
small figuressummary() and plot_effect() flag RIT and
RID values that should not be interpreted- Clearer printed output
- Dependencies on dplyr and purrr removed
Documentation
- Help pages were merged into a compact set:
?rmedsem,
?rmedsem-methods, ?effect-sizes and
?plot.rmedsem
- Examples for all backends
- Corrected and completed the documentation of the datasets
(incl. sources and response scales)
- New article “Working with the results”
- Added CRAN badge to README
rmedsem 1.0.0
- Initial CRAN release.
- Support for mediation analysis with lavaan (CB-SEM), cSEM (PLS-SEM),
blavaan (Bayesian SEM), and modsem (moderated mediation).
- Implements Baron and Kenny (1986) and Zhao, Lynch & Chen (2010)
approaches.
- Reports indirect effects via Sobel, Delta, and Monte-Carlo
methods.
- Effect size measures: RIT and RID.
- Summary and plot methods for results.
- Fix standard error calculation for standardized coefficients in
rmedsem.lavaan and rmedsem.modsem (contributed by Kjell Slupphaug).
- New
ci.two.tailed parameter for configurable confidence
interval width.