- Added reference to Lindmark (2022).
- Added a clarification on the types of exposure variables allowed to
the help-file of
sensmediation.
- Added missing package anchors for Rd targets.
- Fixed a bug in the
calc.effects function that could
affect the estimated direct and indirect effects with moderation on more
than one covariate if the exposure-covariate, mediator-covariate,
exposure-mediator-covariate interaction terms were not in the same order
as the main effects of the covariates
- Fixed an error in the log-likelihoods for cases with probit
model.expl and continuous model.resp as well
as cases with probit model.resp and continuous
model.expl which gave results for -rho instead of rho and
vice versa.
- Corrected warning messages for omitted
exp.name and
med.name arguments in sensmediation
- Estimated outcome and mediator model parameters separated into
different categories depending on effect type are now stored in
effectsMed objects
- The partial derivatives of the NDE, NIE and TE wrt the mediator and
outcome model parameters for each value of Rho are now stored in
effectsMed objects
- A list with the joint covariance matrix of the outcome and mediator
model parameters for each value of Rho is now stored in
effectsMed objects
- A data set consisting of a subsample from Riksstroke, the Swedish
Stroke Register, is now provided and used to exemplify the package
functions.
- The covariate names in the list of the
covariates
argument now can (need to) be given as coded in the original data rather
than the way they are recoded in the glm output.
- A note is printed if one or more covariates in
covariates are not found in the mediator and outcome models
and therefore not conditioned on. Variables not conditioned on are not
printed in the output from summary.effectsMed.
- Optimization function changed from
optimx to
maxLik
- Analytic Hessians of the log-likelihoods have been implemented
- Default optimization method changed from BFGS to Newton Raphson
- Arguments allowed to be passed to the maximization function
(
method and control)
- The
sensmediation argument covariance has
been deprecated
- The
sensmediation arguments out.full,
med.full and all.interactions have been
deprecated
- New
sensmediation arguments: exp.name and
med.name
- Errors in the analytic gradients of the cases with probit
model.expl and continuous model.resp as well
as cases with probit model.resp and continuous
model.expl have been fixed.