The complexity of conducting regression analysis over multiple covariates is well-documented. The challenge only intensifies when coupled with small sample sizes or missing data sets. LogXact aims to provide simple and accurate solutions for such dificulties. Users performing analysis on binary data, count data and all types of categorical data benefit from LogXact’s tools for exact inference, Monte Carlo methods, MCMC, and Firth’s procedures. LogXact’s powerful algorithms are built to analyze stratified and unstratified data sets of all sizes, and models are provided for cases of missing categorical (discrete) data.
LogXact performs regression analysis for continuous, binary, polytonomous and count data. It can also apply advanced regression techniques to data sets with missing values.
Only in LogXact can users accurately fit general linear models in cases of missing categorical covariates (models include Logit, Probit, CLoglog, Poisson and Normal).
LogXact provides options to handle large data sets using exact methods, Monte Carlo sampling and Markhov Chain Monte Carlo sampling. A useful ‘Exploration Mode’ allows users to specify parameters to build networks that satisfy analysis needs.
Saddle point approximation method for inference conditioned on more than one nuisance parameters using higher-order asymptotic. Bias corrected Maximum Likelihood Estimation (MLE) for multinomial data. Bias corrected MLE for Exact Poisson regression to complement bias correction method by Firth.