A new variable selection framework is applied to LUCID. * A lasso type penalty is applied on the mean of biomarkers * A glasso method is applied on the variance-covariance structure to achieve sparsity covariance matrix * We apply a new variable selection criteria, which takes both mean and coviarnce matrix of biomarkers into account.
pred.lucid(). Now it can predict both latent cluster and the outcome.This is a feature update to the whole package. It rewrite all the codes to make the model fitting procedure much faster (10 to 50 times) than v1.0.0. Also, the grammar of LUCID changed to a more user-friendly version. (Please note, this version is not backward compatible)
est.lucid(): previously called est_lucid. Fit the LUCID mode much faster; use mclust to initialize and produce a more stable estimate of the model; fix the bugs dealing with missing values in biomarker data.summary.lucid(): previously called summary_lucid. An S3 method function which can directly be called by summary; provide with a nice table with detailed interpretation of the model; add option to calculate 95% CI based on result returned by boot.lucid().plot.lucid(): previously called plot_lucid. An S3 method function which can be directly called by plot; change the color palette.predict.lucid(): previously called pred_lucid. An S3 method function which can be directly called by predict.boot.lucid(): previously called boot_lucid. Provide with a neat output.