EM_fit: EM Fit Of K-mixture Components

In sp2019-antibiotics/semiparclust: ECOFF Using Semi-Parametric Clustering

Description

For the estimation of GMM with only K=1 component we have used the function MASS::fitdistr().For the estimation of GMM with k=2,...,K components an EM algorithm is used to obtain needed output.

Usage

EM_fit(data, K, take_resist = F, plot = F, tolerance = 0.001)

Arguments

data	vector of binned data
K	number of Gaussian components for mixture model.
take_resist	logical, whether or not to consider first bin referring to resistant observations, default is FALSE.
plot	logical, whether or not to plot final fit of GMM on original data, default is FALSE.
tolerance	tolerance for the Aitken-acceleration based stopping criterion, default is 0.001.

Value

• pi: mixing parameters pik of final GMM

• mu: means muk of the final GMM

• sigma: standard deviation sigma of the final homoscedastic GMM

• loglik: vector of log-likelihood values during the iterative procedure - in case of K=1 only last value of the log-likelihood will be returned

• AIC: AIC of the GMM fitted to the given data

• BIC: BIC of the GMM fitted to the given data

• data_used: data which were used for the fitting procedure

• reconstr_data: reconstructed data of the given binned data

• bins: bins

sp2019-antibiotics/semiparclust documentation built on Nov. 5, 2019, 9:14 a.m.