0

I have a design in which I have pairs of individuals that compete against each other. Each pair thus has one winner.

  • A vs B B wins
  • B vs C C wins
  • ...

I have also measured a continuous variable (X) of each individual and I expect that X predicts wins. Normally I would thus calculate a difference score for each pair and then predict the win E.g.,

model <- glm(data$Winner ~ data$diff, family = "binomial" ) summary(model)

However, I do have a few individuals in the dataset who appear in multiple pairs (hence an "incomplete round-robin"):

  • A vs B
  • C vs D
  • A vs E
  • F vs G
  • ...

I don't know how to account for these "overlapping" pairs in the analyses.

Thanks a lot for any input.

xoph09
  • 1

1 Answers1

0

You could also select a discrete choice model, where you model "A vs B" as a situation where A and B are possible choices and if A wins against B you model A as the chosen alternative. In R I know the package mlogit that can calculate such models.

maxatSOflow
  • 269
  • 3
  • 10