Examples

Cytel's LogXact 5 vs. SAS's PROC LOGISTIC:
Racial Bias in Death Sentencing in New Jersey

(Source: Office of the N.J. Public Defender)

This study was conducted to determine whether race was a factor in applying the death penalty in New Jersey between 1982 and 1998.
(160 observations, 10 covariates.)

A model was fit with the following 10 covariates:

  BLACKD: 1 = black defendant
  OTHAGG: 1 = the presence of at least one of a number of factors
              (assignment of blame, mutilation, threats and suffering)
POOLADEH: 1 = the presence of at least one of a number of aggravating factors
              (prior murder, received payment, etc.)
 POOLGBF: 1 = the presence of at least one of several factors
              (the committing of a contemporaneous felony, grave risk to another person,
              and act committed while escaping)
 POOLMCG: 1 = the presence of at least one of several factors
POOMABEF: 1 = the presence of at least one of several factors
   SESF1
  V4CPTY: 1 = outrageously wanton or vile act
  V5DBTY: 1 = the presence of mental disease or intoxication
WHITEVIC: 1 = white victim

The primary question is whether the defendant's race (BLACKD covariate) is a significant factor in the application of the death penalty, while adjusting for the effects of the other confounders. The results show that this is indeed a significant factor. Note also that SAS's PROC-LOGISTIC cannot solve the problem; only LOGXACT 5 can do it.

LOGXACT 5 RESULTS

=============================================================================================
Hypothesis Testing
Tests 
Type of test           Statistic DF        P Value   P-Mid     Std.Err   
=============================================================================================

Likelihood ratio       4.7903    1         0.0286    NA        
Wald                   4.5395    1         0.0331    NA        
Score                  4.7087    1         0.0300    NA    
    
Exact Likelihood ratio:Monte
                       4.7903    NA        0.0510    NA        0.0070    
( Monte Carlo estimate : seed = 1038330779,samples = 1000 )

=============================================================================================

Analysis time: 00:15:29 
=============================================================================================


SAS's PROC LOGISTIC RESULTS
WARNING: There is not enough memory available for exact computations.

LOGXACT 5--
Download LogXact data
View Instructions for LogXact 5 analysis

PROC LOGISTIC--
Download SAS code (includes data)