What is LogXact® for?

Cytel, the acknowledged leader in exact statistical methods, helped pioneer exact methods for binary logistic regression and multinomial regression. First introduced in 1987, LogXact is unequivocally the fastest and most powerful logistic regression analysis software available today.

Why is it important?

Even with small sample sizes and missing data, LogXact® users enjoy complete confidence in their logistic regression analyses.

LogXact is the only tool to provide these major logistic regression modeling developments:

• Only in LogXact, Firth's Penalized Maximum Likelihood Estimate (Biometrika [1993], 80, 1, pp27 - 38) corrects for separation bias, while exhibiting lower mean square error than usual Maximum Likelihood (ML) estimations click for example
• Fitting GLM in the presence of missing categorical covariates. With missing discrete covariates, LogXact employs Chen and Ibrahim's EM algorithm to produce GLM parameter estimates for the covariates. Logit, Probit, CLogLog, Poisson and Normal links are all supported

What's new in the latest version?

Windows® 7 and Vista support

Cytel works continuously to allow our customers a choice of supported operating systems. Cytel's StatXact® and LogXact® are now both approved for use in Windows® 7 and Windows® Vista.
See the complete Supported OS, at right.

R integration

Users can now use R scripts in conjunction with Cytel’s exact statistics software. Computers with R 2.3 or newer may run existing R files or write and run R programs with the R output displayed in LogXact®.

R programs can also run analyses of either StatXact® or LogXact® datasets. All R outputs are displayed in the Cytel Studio interface (except R plots) - see example here.

In addition to R, Cytel continues support of its own batch language to for custom analysis and regression calculations automation.

Latest validated methods include

• Best subset selection in binary logistic regression
• Force inclusion of variables to the best subset
• Profile Likelihood Confidence Intervals for parameters of binary logistic regression
• Firth's correction to Profile Likelihood Confidence Intervals
  See Penalized Maximum Likelihood Method (PMLE) example of Firth's correction, the procedure for obtaining the Profile Likelihood-based confidence intervals for the parameters with the Venzon and Moolgavkar algorithm (1988)).

See examples

See LogXact examples using real-world data sets, or the LogXact brochure

Crossover™ trial tool free

As a bonus, Cytel’s Crossover™ module for designing trials and analyzing crossover data is free for StatXact 9 or LogXact 9 registered users.

Questions?

Call Connor Tubridy at +1.617.528.7121 or email sales@cytel.com

StatXact® 9 and the latest LogXact® now feature the complete user manual online within the software.