Sorry I should’ve started with my overall problem statement.
I’m trying to model the points scored in an American football game. Because these points are scored only in increments of 3 and 7 typically (6, 8 and 2 occasionally, but not very frequently), scoring tends to group around certain numbers far more frequently making Poisson or Negative Binomial regressions a poor choice as I understand. To capture these traits of the data, I chose an Ordered Logistic as my likelihood with 81 cutpoints to start due to 81 being the most points ever scored (inclusive of 0).
