 |
 |
 |
 |
 |
 |
|
LogXact 4 Results for Insurance Fraud Data
1-Sided P-Values for Logistic Regression |
Conclusion |
||
|
|
|
P-value |
LogXact 4 reveals that there is a major difference between p-values produced by exact logistic regression and asymptotic logistic regression. Moreover, only LogXact 4 with its ground-breaking hybrid network-Monte Carlo algorithm can fit an exact logistic regression to this data set with its sizable number of individual covariates. (Even prior versions of LogXact 2 cannot do it!)
Data from the Automobile Insurance Bureau of Massachusetts, cited in Mehta, Patel and Senchaudhuri (JASA, vol 95, no 449, March 2000: Efficient Monte Carlo Methods for Conditional Logistic Regression). |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Note the substantial over-estimation of significance by the asymptotic method, as compared to the exact method. | |||
|
||