data	The dataset (including the covariate values).
timeintervals	The name of the column containing the intervals between measurement occasions. The default is NULL, which means that the measurement occasions are assumed to be equidistant.
identifier	The name of the column containing the subject identifiers.
n_state	The number of states that was used for the estimation with step1().
postprobs	The posterior state-membership probabilities of step2().
transitionCovariates	The covariate(s) for the transition intensities. The default is NULL, which means that no covariate effects are estimated.
initialCovariates	The covariate(s) for the initial state probabilities. The default is NULL, which means that no covariate effects are estimated.
n_starts	The number of random start sets (for details, see Supplementary Material S.5.4). The default is 25.
n_initial_ite	The number of initial iterations that should be performed for each start set. The default is 10.
method	The estimation method. The default is "BFGS", which is usually faster and more stable when including covariates. The alternative is "CG".
max_iterations	The maximum number of iterations after which the estimation stops regardless of whether convergence has been reached or not. The default is 1000.
tolerance	The convergence tolerance (for details, see Supplementary Material S.5.3). The default is 1e-10. When the message occurs that the model has likely not converged because the Hessian is not positive definite, it is advisable to set the argument to a lower value and repeat the analysis (e.g., 1e-16; Jackson, 2011).
scaling	A scaling parameter for the loglikelihood that can prevent numerical problems from occurring, which is internally passed to the optimization function optim(). An appropriate scale value is close to -2 times the loglikelihood, but the loglikelihood is of course unknown prior to estimating the model. Therefore, by default, lmfa uses an approximation, which is based on the loglikelihood value of a CT-LMM without transitions. Next to this default (i.e., scaling = "proxi"), it is also possible to specify own scale values.