A Multivariate Extension of the Dynamic Logit Model for Longitudinal Data Based on a Latent Markov Heterogeneity Structure
Citations
78 citations
Cites background or methods from "A Multivariate Extension of the Dyn..."
...The resulting model is a latent Markov regression model (Bartolucci and Farcomeni 2009)....
[...]
...Bartolucci and Farcomeni (2009), in a different context, report on bias in approximating a latent Gaussian distribution with a similar finite mixture structure....
[...]
...See also Dardanoni and Forcina (1998) and Bartolucci and Farcomeni (2009)....
[...]
...Keywords Asymmetric Laplace distribution · Hidden Markov model · Longitudinal data · Quantile regression...
[...]
...In the latent Markov context, BIC is usually preferred since AIC often leads to overestimation of the number of latent states (see for instance a brief simulation study in Bartolucci and Farcomeni 2009, and Boucheron and Gassiat 2007 for a more general discussion)....
[...]
61 citations
Cites background or methods from "A Multivariate Extension of the Dyn..."
...We refer to Bartolucci (2006) and Bartolucci and Farcomeni (2009) for details....
[...]
...The multivariate LM model with covariates affecting the manifest probabilities proposed by Bartolucci and Farcomeni (2009) was applied by these authors to data extracted from the “Panel Study of Income Dynamics” database (University of Michigan)....
[...]
...The more interesting methods are based on the information matrix obtained from the EM algorithm by the technique of Louis (1982) or related techniques; see for instance Turner et al. (1998) and Bartolucci and Farcomeni (2009)....
[...]
...The formulation proposed by Bartolucci and Farcomeni (2009) includes the covariates in the measurement part of the model in the presence of multivariate responses....
[...]
...Bartolucci and Farcomeni (2009) used an LM model with covariates affecting the manifest probabilities since they were interested in separately estimating the effect of each covariate on each outcome....
[...]
48 citations
Cites background or methods from "A Multivariate Extension of the Dyn..."
...In any case, this assumption can be easily relaxed when all outcomes are categorical (Bartolucci and Farcomeni, 2009), using a marginal parameterization based on logits and log-odds ratios....
[...]
...We also pay attention to the computation of the standard errors for the parameter estimates by employing a method proposed in Bartolucci and Farcomeni (2009). The remainder of the article is organized as follows....
[...]
..., Creemers et al., 2010; Viviani, Rizopoulos, and Alfó, 2014) consider shared-parameter models as a separate framework. Selection models for discrete longitudinal data have been proposed, among others, by Molenberghs, Kenward, and Lesaffre (1997) and Ten Have et al. (1998). The advantages of shared-parameter models like the one we propose are that a directly interpretable marginal model is obtained for the observed outcomes, even if at the price of a heavier estimation procedure....
[...]
...We also pay attention to the computation of the standard errors for the parameter estimates by employing a method proposed in Bartolucci and Farcomeni (2009)....
[...]
44 citations
Cites methods from "A Multivariate Extension of the Dyn..."
...Then, maximum likelihood (ML) estimation may be performed by an adaptation of the EM algorithm for the LM model that was described by Bartolucci and Farcomeni (2009); see also Baum et al. (1970) and Dempster et al. (1977)....
[...]
...The NR algorithm is based on the observed information matrix which is obtained by the same numerical method as proposed by Bartolucci and Farcomeni (2009)....
[...]
...Following Bartolucci and Farcomeni (2009), the score vector is computed as the first derivative of the expected value of the completedata log-likelihood, which is obtained after an E-step....
[...]
...This is a formulation of LM type, which was employed by Bartolucci and Farcomeni (2009) to propose a flexible class of models for multivariate categorical longitudinal data....
[...]
...These posterior probabilities may be computed by suitable recursions; see Baum et al. (1970) and Bartolucci and Farcomeni (2009) for details....
[...]
44 citations
Cites background or methods from "A Multivariate Extension of the Dyn..."
...For multivariate continuous data, attention is commonly focused on Gaussian HMMs (Bartolucci and Farcomeni 2010; Volant et al. 2014; Holzmann and Schwaiger 2015), with few notable exceptions (Bartolucci and Farcomeni 2009; Bulla et al. 2012; Lagona, Maruotti, and Padovano 2015)....
[...]
..., 2014; Bartolucci and Farcomeni, 2010), with few notable exceptions (Lagona et al., 2015; Bulla et al., 2012; Bartolucci and Farcomeni, 2009)....
[...]
References
8,258 citations
8,095 citations
6,804 citations
"A Multivariate Extension of the Dyn..." refers methods in this paper
...Finally, we deal with prediction of the vector of responses and illustrate the Viterbi algorithm (Viterbi 1967; Juang and Rabiner 1991) for path prediction, i.e., prediction of the sequence of latent states of a given subject on the basis of his/ her observable covariates and response variables....
[...]
...To predict the entire sequence of latent states, we can use the Viterbi algorithm (Viterbi 1967; Juang and Rabiner 1991)....
[...]
...Finally, we deal with prediction of the vector of responses and illustrate the Viterbi algorithm (Viterbi 1967; Juang and Rabiner 1991) for path prediction, i....
[...]
6,234 citations