StatMixHMM-class function

A Reference Class which contains statistics of a mixture of HMM model.

StatMixHMM contains all the statistics associated to a MixHMM

model, in particular the E-Step of the EM algorithm. class

Fields

• tau_ik: Matrix of size $(n, K)$ giving the posterior probabilities that the curve $y_{i}$ originates from the $k$-th HMM model.

• gamma_ikjr: Array of size $(nm, R, K)$ giving the posterior probabilities that the observation $y_{ij}$

   originates from the $r$-th regime of the $k$-th HMM model.
  

• loglik: Numeric. Log-likelihood of the MixHMM model.

• stored_loglik: Numeric vector. Stored values of the log-likelihood at each iteration of the EM algorithm.

• klas: Row matrix of the labels issued from tau_ik. Its elements are $klas[i] = z_i$, $i = 1,\dots,n$.

• z_ik: Hard segmentation logical matrix of dimension $(n, K)$

   obtained by the Maximum a posteriori (MAP) rule: c("$z_ik = 1 if z_i = arg max_k P(z_{ik} = 1 | y_{i};\n$", "$  \\Psi) = tau_ik; 0 otherwise$").
  

• smoothed: Matrix of size $(m, K)$ giving the smoothed time series. The smoothed time series are computed by combining the time series $y_{i}$ with both the estimated posterior regime probabilities gamma_ikjr and the corresponding estimated posterior cluster probability tau_ik. The k-th column gives the estimated mean series of cluster k.

• BIC: Numeric. Value of BIC (Bayesian Information Criterion).

• AIC: Numeric. Value of AIC (Akaike Information Criterion).

• ICL1: Numeric. Value of ICL (Integrated Completed Likelihood Criterion).

• log_alpha_k_fyi: Private. Only defined for calculations.

• exp_num_trans: Private. Only defined for calculations.

• exp_num_trans_from_l: Private. Only defined for calculations.

Methods

• computeStats(paramMixHMM): Method used in the EM algorithm to compute statistics based on parameters provided by the object paramMixHMM of class ParamMixHMM .

• EStep(paramMixHMM): Method used in the EM algorithm to update statistics based on parameters provided by the object paramMixHMM of class ParamMixHMM

   (prior and posterior probabilities).
  

• MAP(): MAP calculates values of the fields z_ik and klas

   by applying the Maximum A Posteriori Bayes allocation rule.
   
    c("$z_ik = 1 if\n$", "$      z_i = arg max_k P(z_{ik} = 1 | y_{i};\\Psi) = tau_ik; 0 otherwise$").
  

See Also

ParamMixHMM