rlacalc 0.4.0

Statistical calculations for risk-limiting election tabulation audits (RLAs)

rlacalc crunches the numbers for risk-limiting post-election audits. It can calculate sample sizes, as well as calculate risk levels based on observed ballots. It can handle many of the commonly used RLA tests, including ballot-level comparison via Kaplan-Markov and ballot polling via BRAVO.

Goals

It is important to emphasize that the output of an audit should be a report providing evidence related to election outcome, details about and explanations of any discrepancies found, and conclusions based on that evidence. See the paper Evidence-Based Elections Stark and Wagner and the report Risk-Limiting Post-Election Audits: Why and How.

rlacalc helps observers to explore the best way to do an audit beforehand, and to check the numbers in reports published after an audit, e.g. as described for a Public RLA Oversight Protocol.

rlacalc can be used to perform calculations from the command line or used as a library in Python. In addition, if the hug package is available, it can be accessed over the web.

Installation

rlacalc works with Python 2.7 and Python 3.x. There are no other mandatory dependencies beyond the standard library.

If hug is installed, rlacalc can be deployed via a web server.

Usage

$ rlacalc --help
...

$ rlacalc --test
...

$ rlacalc -m 1 -n
Sample size = 479 for margin 1%, risk 10%, gamma 1.03905, o1 0, o2 0, u1 0, u2 0

$ rlacalc -m 1 -n --o1 1
Sample size = 615 for margin 1%, risk 10%, gamma 1.03905, o1 1, o2 0, u1 0, u2 0

$ rlacalc -m 3 -r 5
KM_exp_rnd  = 226 for margin 3%, risk 5%, gamma 1.03905, or1 0.001, or2 0.0001, ur1 0.001, ur2 0.0001, roundUp1 1, roundUp2 0

Background

For tools implemented in javascript, with plots, see

• Tools for Comparison Risk-Limiting Election Audits
• Tools for Ballot-Polling Risk-Limiting Election Audits

History

rlacalc was included as part of audit_cvrs in 2015 to help run some early pilot audits in Colorado as part of the The Colorado Risk-Limiting Audit Project (CORLA).