I notice that the genotype Explanation is generated separately from the way the actual LR is generated, and it has gotten out of synch in a few edge cases. It seems better to return an object such as LrWithExplanation that includes the likelihood ratio as well as a score. There was also a bug with returning a negative LR that led to an NaN error on some edge cases. I have fixed this and will implement a new class called GenotypeLrWithExplanation.
I fixed this, including a bug with how the genotype LR is calculated in an edge case, and it seems to have slightly improved the performance on our test set of 384 cases as well!
I notice that the genotype Explanation is generated separately from the way the actual LR is generated, and it has gotten out of synch in a few edge cases. It seems better to return an object such as LrWithExplanation that includes the likelihood ratio as well as a score. There was also a bug with returning a negative LR that led to an NaN error on some edge cases. I have fixed this and will implement a new class called GenotypeLrWithExplanation.