SaemixRes-class function

Class "SaemixRes"

An object of the SaemixRes class, representing the results of a fit through the SAEM algorithm. class

Slots

• modeltype: string giving the type of model used for analysis
• status: string indicating whether a model has been run successfully; set to "empty" at initialisation, used to pass on error messages or fit status
• name.fixed: a vector containing the names of the fixed parameters in the model
• name.random: a vector containing the names of the random parameters in the model
• name.sigma: a vector containing the names of the parameters of the residual error model
• npar.est: the number of parameters estimated (fixed, random and residual)
• nbeta.random: the number of estimated fixed effects for the random parameters in the model
• nbeta.fixed: the number of estimated fixed effects for the non random parameters in the model
• fixed.effects: a vector giving the estimated h(mu) and betas
• fixed.psi: a vector giving the estimated h(mu)
• betas: a vector giving the estimated mu
• betaC: a vector with the estimates of the fixed effects for covariates
• omega: the estimated variance-covariance matrix
• respar: the estimated parameters of the residual error model
• fim: the Fisher information matrix
• se.fixed: a vector giving the estimated standard errors of estimation for the fixed effect parameters
• se.omega: a vector giving the estimated standard errors of estimation for Omega
• se.cov: a matrix giving the estimated SE for each term of the covariance matrix (diagonal elements represent the SE on the variances of the random effects and off-diagonal elements represent the SE on the covariance terms)
• se.respar: a vector giving the estimated standard errors of estimation for the parameters of the residual variability
• conf.int: a dataframe containing the estimated parameters, their estimation error (SE), coefficient of variation (CV), and the associated confidence intervals; the variabilities for the random effects are presented first as estimated (variances) then converted to standard deviations (SD), and the correlations are computed. For SD and correlations, the SE are estimated via the delta-method
• parpop: a matrix tracking the estimates of the population parameters at each iteration
• allpar: a matrix tracking the estimates of all the parameters (including covariate effects) at each iteration
• indx.fix: the index of the fixed parameters (used in the estimation algorithm)
• indx.cov: the index of the covariance parameters (used in the estimation algorithm)
• indx.omega: the index of the random effect parameters (used in the estimation algorithm)
• indx.res: the index of the residual error model parameters (used in the estimation algorithm)
• MCOV: a matrix of covariates (used in the estimation algorithm)
• cond.mean.phi: a matrix giving the conditional mean estimates of phi (estimated as the mean of the conditional distribution)
• cond.mean.psi: a matrix giving the conditional mean estimates of psi (h(cond.mean.phi))
• cond.var.phi: a matrix giving the variance on the conditional mean estimates of phi (estimated as the variance of the conditional distribution)
• cond.mean.eta: a matrix giving the conditional mean estimates of the random effect eta
• cond.shrinkage: a vector giving the shrinkage on the conditional mean estimates of eta
• mean.phi: a matrix giving the population estimate (Ci*mu) including covariate effects, for each subject
• map.psi: a dataframe giving the MAP estimates of individual parameters
• map.phi: a dataframe giving the MAP estimates of individual phi
• map.eta: a matrix giving the individual estimates of the random effects corresponding to the MAP estimates
• map.shrinkage: a vector giving the shrinkage on the MAP estimates of eta
• phi: individual parameters, estimated at the end of the estimation process as the average over the chains of the individual parameters sampled during the successive E-steps
• psi.samp: a three-dimensional array with samples of psi from the conditional distribution
• phi.samp: a three-dimensional array with samples of phi from the conditional distribution
• phi.samp.var: a three-dimensional array with the variance of phi
• ll.lin: log-likelihood computed by lineariation
• aic.lin: Akaike Information Criterion computed by linearisation
• bic.lin: Bayesian Information Criterion computed by linearisation
• bic.covariate.lin: Specific Bayesian Information Criterion for covariate selection computed by linearisation
• ll.is: log-likelihood computed by Importance Sampling
• aic.is: Akaike Information Criterion computed by Importance Sampling
• bic.is: Bayesian Information Criterion computed by Importance Sampling
• bic.covariate.is: Specific Bayesian Information Criterion for covariate selection computed by Importance Sampling
• LL: a vector giving the conditional log-likelihood at each iteration of the algorithm
• ll.gq: log-likelihood computed by Gaussian Quadrature
• aic.gq: Akaike Information Criterion computed by Gaussian Quadrature
• bic.gq: Bayesian Information Criterion computed by Gaussian Quadrature
• bic.covariate.gq: Specific Bayesian Information Criterion for covariate selection computed by Gaussian Quadrature
• predictions: a data frame containing all the predictions and residuals in a table format
• ppred: a vector giving the population predictions obtained with the population estimates
• ypred: a vector giving the mean population predictions
• ipred: a vector giving the individual predictions obtained with the MAP estimates
• icpred: a vector giving the individual predictions obtained with the conditional estimates
• ires: a vector giving the individual residuals obtained with the MAP estimates
• iwres: a vector giving the individual weighted residuals obtained with the MAP estimates
• icwres: a vector giving the individual weighted residuals obtained with the conditional estimates
• wres: a vector giving the population weighted residuals
• npde: a vector giving the normalised prediction distribution errors
• pd: a vector giving the prediction discrepancies

Objects from the Class

An object of the SaemixData class can be created by using the function saemixData and contain the following slots:

Methods

• [<-: signature(x = "SaemixRes"): replace elements of object
• [: signature(x = "SaemixRes"): access elements of object
• initialize: signature(.Object = "SaemixRes"): internal function to initialise object, not to be used
• print: signature(x = "SaemixRes"): prints details about the object (more extensive than show)
• read: signature(object = "SaemixRes"): internal function, not to be used
• showall: signature(object = "SaemixRes"): shows all the elements in the object
• show: signature(object = "SaemixRes"): prints details about the object
• summary: signature(object = "SaemixRes"): summary of the results. Returns a list with a number of elements extracted from the results ().

Examples

methods(class="SaemixRes")

showClass("SaemixRes")

References

E Comets, A Lavenu, M Lavielle M (2017). Parameter estimation in nonlinear mixed effect models using saemix, an R implementation of the SAEM algorithm. Journal of Statistical Software, 80(3):1-41.

E Kuhn, M Lavielle (2005). Maximum likelihood estimation in nonlinear mixed effects models. Computational Statistics and Data Analysis, 49(4):1020-1038.

E Comets, A Lavenu, M Lavielle (2011). SAEMIX, an R version of the SAEM algorithm. 20th meeting of the Population Approach Group in Europe, Athens, Greece, Abstr 2173.

See Also

saemixData SaemixModel saemixControl saemix

Author(s)

Emmanuelle Comets emmanuelle.comets@inserm.fr

Audrey Lavenu

Marc Lavielle.