A number of projects have each been graded by a number of judges. Not all projects were graded by all judges. In general each judge provides grades for a subset of projects and each project has a small number of judges.
We are interested in correcting for any bias displayed by particular judges. That is, we want to estimate a model where each grade provided by a judge is a random variable centered at a value that is the true value of the project plus a bias by the judge.
All calculations are currently in Jupyter notebooks. This was done for a specific class, so you can see the process on the real grades.
The frist generates two csv files and the second generates and .xlsx showing the final results