wjiang94 / misaem

package R

misaem package

Introduction

misaem is a package to perform linear regression and logistic regression with missing data, under MCAR (Missing completely at random) and MAR (Missing at random) mechanisms. The covariates are assumed to be continuous variables. The methodology implemented is based on maximization of the observed likelihood using EM-types of algorithms. The package includes:

1. Parameters estimation.
2. Estimation of standard deviation for estimated parameters.
3. Model selection procedure based on BIC.

Installation of package

Now you can install the package misaem from CRAN.

install.packages("misaem")

Using the misaem package

Basically,

1. miss.glm is the main function performing logistic regression with missing values.
2. miss.lm is the main function performing linear regression with missing values.

For more details, You can find the vignette, which illustrate the basic and further usage of misaem package:

library(misaem)
vignette('misaem')

Reference

Logistic Regression with Missing Covariates -- Parameter Estimation, Model Selection and Prediction (2020, Jiang W., Josse J., Lavielle M., TraumaBase Group), Computational Statistics & Data Analysis.

misaem's Issues

Undefined variable `p` in miss.lm.fit function

Hi Wei,
I have a small question/issue for the miss.lm.fit function: when trying to use it (via the miss.lm function), I get an error message saying that the variable p isn't defined.
And indeed, when I looked into the miss.lm.fit function, I found that there is a condition using this variable p (line 32):
if(any(apply(is.na(x),1,sum)==p)) prior to its definition.

Display Z-value and odds ratio.

Hi,
I'am replicating a logistical model that was made on SAS and I used misaem to avoid deleting the rows with missing values.

I have a first question about the specificty of the package. The description on Cran was a bit technical for me... Basically, does the package allows to make logistical regressions with NA or does is compute something different that regular logistical models with NA on other environnement (Sas, stata, python etc)?

My second question is about the possibility to display Z-value and odds ratios? I tried to use the glm fonction "odds.ratio" but it isn't compatible. How can theses values be found with a miseam object?

Thank you for this package!
Best,
Victor