Estimate parameters of linear regression and logistic regression with missing covariates with missing data, perform model selection and prediction, using EM-type algorithms.
misaem is an implementation of methodology which performs statistical inference for logistic regression model with missing data. This method is based on likelihood, including:
Now you can install the package misaem from CRAN.
miss.saemcontains the procedure of estimation for parameters, as well as their variance, and observed likelihood.
model_selectionaims at selecting a best model according to BIC.
pred_saemperforms prediction on a test set which may contain missing values.
For more details, You can find the vignette, which illustrate the basic and further usage of misaem package:
Stochastic Approximation EM for Logistic regression with missing values (2018, Jiang W., Josse J., Lavielle M., Traumabase group)" arxiv link.
A minor release mainly fixing bugs and typos:
Fix a bug in
model_selection, now it can correctly perform model selection if the full model is the best model.
pred_saem, two methods for prediction of test set with missingness are provided.
Fix some typos in the vignettes. The length of codes now fits the page wide of html.
Delete unused Imports ‘magrittr’ in DESCRIPTION file.
Change the index of vignitte from 'SAEM' to 'misaem tutorial'.