Bandit Release Notes

• Enabled banding of 'Long' files (-i R)
• Automatically detects numbers that should be categoral data (e.g. phone numbers, social security numbers
  • The command (-g 2) will enable this feature
  • The command (-g 1) will skip this logic, but will designate single integers as categoral
• Bandit can now prune categories that share little mutual information with the dependent category (-p 1)
  • These categories will NOT appear in the banded file, but will appear in the breaks and mi files

Enhanced Functionality

• Bandit will now round breakpoints of bands to the nearest 2 decimal places
• Now sampling up to 10,000,000 lines when banding large files
• Improvements made with respect to file size and estimations of the number of rows in a data file
• Improved speed on Windows machines
• Better 'bad' data detection by the flaging of “silly” numbers as missing data

Defects Resolved

• Bandit for Red Hat and Windows will now place the correct bracket labels around bins
• Escaped quotes can now be handled, and the missing field issue has been resolved
• File headers will no longer cause errors due to skipping the line ending
• Corrected instances of auto bands being created with size of zero
• Columns containing percent sign are now banding correctly