New version of the vignette.
seqaddNA()now has apcont` argument, which allows to
modify the probability that a gap with missing data will be
continued.
Bug regarding the frame.radius parameter that was always set as frame.radius=0, whatever the user provided corrected.
Bug regarding edge case with only one level in the dependent variable corrected.
For parallel computing, the maximum number of cores is set as availableCores()-1, instead as detectCores()-1.
Warnings now appear when the number of specified cores exceeds the number of multiple imputations, or the number of availableCores()-1.
Explanations of seqaddNA() have been
deepened.
Explanations of seqimpute() regarding the slots of
the seqimp object that is returned have been deepened.
The colors for the seqmissIplot() and
seqmissfplot plots have been changed to improve
readability.
Each variable of the imputed object now have the same levels.
The quality and speed of the code were increased.
Reference to new article about MICT-timing added.
seqimpute() now has an argument
end.impute argument, which specifies if missing data at the
end of sequences should be imputed or not.
seqmissIplot(), seqmissfplot() and
seqmissimplic now have an argument void.miss,
which specifies whether the void values of a provided object of class
stslist should be considered as missings or not.
seqaddNA() now has an argument maxprop,
which specifies the maximum proportion of missing data that is allowed
to be simulated in a sequence.
Fixes issues with seqimpute() when more than one row
only with NA’s end the dataset to impute.
In seqmissIplot() and seqmissfplot(),
the states in the plots are now labeled as ‘missing’ and ‘not missing’
instead of ‘missing’ and ‘observed’ to account for uneven sequence
length.
Fixes issues in seqimpute() related to the
preparation of the data when an object of class stslist,
built with the TraMineR package is provided.
seqimpute() now returns an object of class
seqimp. In particular, the include and
mice.return arguments are no longer relevant and have been
removed.
The OD argument has been renamed to
data. The argument OD itself is
deprecated.
The CO argument has been renamed to
covariates. The argument CO itself is
deprecated.
The COt argument has been renamed to
time.covariates. The argument COt itself is
deprecated.
The mi argument has been renamed to m.
The argument mi itself is deprecated.
The dataset provided in the package used to be divided into three
parts: the trajectories (OD), the covariates
(CO), and the time-varying covariates (COt).
They now appear as a single dataset, called
gameadd.
seqimpute() no longer implements linear and ordinal
regressions.
The default argument of m has been set to
5.
seqimpute() implements the MICT-timing imputation
algorithm. The argument timing indicates whether to use
this algorithm or the MICT algorithm, and frame.radius
specifies the radius of the time frame.
The user can now pass a dataset to the seqimpute()
function and specify which columns correspond to the trajectories with
the var argument, to the covariates with the
covariates argument, and the time-varying covariates with
the time.covariates argument.
A vignette has been added.
New seqmissfplot() plot function, which plots the
most frequent patterns of missing data.
New seqmissIplot() plot function, which plots all
patterns of missing data.
New seqmissimplic() function for identifying and
visualizing the states that best characterize sequences with missing
data.
New fromseqimp() function, which converts a
seqimp object into a specified format.
New addcluster() function, which adds a clustering
result to a seqimp object
New seqaddNA() function to simulate missing
data.
New seqcomplete() function, which extracts all
trajectories without missing data.
New seqwithmiss() function, which extracts all the
trajectories with at least one missing value.
seqimpute() now returns an object of class
seqimp. A print, summary, and plot functions have been
added for this object type.
seqTrans() and seqQuickLook() now
accept objects of class stslist built with the
TraMineR package.
A ... argument has been added to
seqimpute() to pass named arguments to the imputation
functions.
Fixes issues in seqimpute() related to the
multinomial model when there is only one state in the dependent
variable.
Fixes issues in seqimpute() related to random forest
when one state does not appear in the dependent variable.
Fixes bug in seqimpute() when a single covariate is
specified.
Fixes bug in seqimpute() related to long internal
gaps.
Fixes bug in seqQuickLook() induced by datasets
without missing data.