4
$\begingroup$

I'm trying to model a pairwise outcome of basketball game scores.

Ie. (94,87),(102,98),(76,54),...

My input variable is a single performance metric for each team.

Ie. (12,9),(14,17),...

Is there a way I can predict the two-part outcome of a game based on the performance metrics?

Thanks!

$\endgroup$
  • $\begingroup$ You could model the two elements in the pair using separate models. Or you could look at the seemingly unrelated regressions (SUR) framework if you think that using the information from the first equation when estimating the second equation, and vice versa, could be useful. (Some say that translating SUR as seemingly related regressions makes more sense than seemingly unrelated regressions, but it is a historical matter.) $\endgroup$ – Richard Hardy Dec 1 '14 at 19:06
4
$\begingroup$

You can certainly have multivariate regression models where the response is not a single value but a vector of possibly dependent values.

You can also have more than one input (multiple regression), so "Team 1 performance" and "Team 2 performance" (or some joint transformation of them, such as their sum and difference, perhaps) could both be predictors in such a model.

So at first glance, multivariate multiple regression would seem to be a good starting point.

The next thing to consider might be whether the response would be better modelled as something other than multivariate normal. You might look into GLMs, for example.

In addition, when looking at pairwise contests like this, it can be difficult to get any absolute measure of individual team performance since they depend heavily on what the other team does - this might lead toward something like Bradley-Terry type models.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.