I am proposing two more courts which can be used for jurors that have skills related to data analysis and statistics. These courts would be appropriate, for example, in situations where freelancers are hired to provide data sets, statistical models, or other statistical services, and jurors must resolve disputes over the quality of the work the freelancer provides.
Data Analysis Court
Court name: Data Analysis
Parent court: General
Fee per juror: .077 ETH
Minimum take: 1990 PNK
alpha: .5
Jurors for jump: 64
timesPerPeriod: [280800, 583200, 583200, 38800]
Description
Required Skills: Jurors should be comfortable dealing with data sets, particularly in being able to understand the context of the data from the structure of the data set, and in estimating what percentage of entries are unusable/would need to be cleaned.
Statistical Modeling Court
Court name: Statistical Modeling
Parent court: Data Analysis
Fee per juror: .235 ETH
Minimum take: 6635 PNK
alpha: .5
Jurors for jump: 32
timesPerPeriod: [367200, 626400, 626400, 604800]
Description
Required Skills: Jurors should be comfortable dealing with data sets. Furthermore, jurors should have a background in statistics equivalent to (at minimum) a university statistics course, and they should be able to analyze the assumptions taken in the creation of statistical models.
Motivation in choices of parameters
In the data analysis court we imagine that the typical juror task will be to look over a submitted data set, which might require 30-45 minutes of effort on average to get a sense of the structure of the data set. In the statistical modeling court we imagine the typical task involving carefully analyzing the assumptions of a submitted statistical model, which might require 1-2 hours of somewhat more specialized effort. In both cases, the minstake is then determined to ensure attack resistance, with reasoning based on that in: https://github.com/kleros/research-docs/blob/master/klerosfeesparameters.pdf The times per period for the data analysis court match those of the general court, while the times per period for the statistical modeling court are somewhat longer to reflect the greater difficulty of the tasks in this court.