I'm working on a project to forecast the distribution of a baseball player's "At Bats per game" (a baseball statistic w/ integer domain) using a player's position in his team's batting order as a predictor (1st up to bat, 2nd up to bat etc).

I went with my initial thought, which was to train a logistic regression on historical data w/ a poisson assumption, plug in the batting order for the game I want to predict and just nab the predicted lambda from the model without actually attempting to predict a value.

I then decided to plot the density of my "fitted" poisson model over the empirical density from my data. Every time I do this the poisson distribution has much more variance than the empirical density. I have several thousand data points so I suspect my empirical density should be pretty good.

I've attached a picture of the empirical density(proportion of all at bats for players with a batting order = 2) plotted against a poisson with lambda predicted for batting order = 2 from my glm. Is there another technique I can use to better predict a distribution of count data? I'm assuming a negative binomial glm will give roughly the same results. I know it's somewhat taboo but it seems like I could get a much better fit for the distribution using a normal distribution from an ols. Thanks for any advice! enter image description here

  • 1
    $\begingroup$ What do you mean by "train a logistic regression ... w/ a poisson assumption"? $\endgroup$ – gung - Reinstate Monica Jun 2 '16 at 2:30
  • $\begingroup$ Oh sorry if that was unclear. I used the glm() function with parameter family = "poisson" is all I mean. Mathematically, im trying to find the value of beta that maximizes the likelihood function (using the poisson pmf) for a given sample. $\endgroup$ – Benjamin Whitesell Jun 2 '16 at 2:32
  • 1
    $\begingroup$ It is not a straightforward counts distribution at all. The number of at bats per game looks to me to be a near constant sum phenomena, and, the number of at bats per player would I think need to be non-independent. So, there is not going to be a right answer to your question, I think. $\endgroup$ – Tim Jun 2 '16 at 3:07
  • $\begingroup$ Would you elaborate a bit on why you suspect non-independence and a constant sum phenomena?Just trying to get a handle on your intuition. I would think at bats should be independent between games so long as I'm looking only at one player per game, clearly at bats between two players on the same team would be correlated but why would they be for a single player? $\endgroup$ – Benjamin Whitesell Jun 2 '16 at 3:41
  • $\begingroup$ Given the rules of baseball, the event a particular player bats in a particular inning is negatively correlated with whether that player batted in the previous inning. Combining that with a low dispersion in the total number of innings, and you get a substantially reduced variance (as you have observed) compared with the Poisson distribution your model assumes. $\endgroup$ – Henry Jun 2 '16 at 6:58

This is probably the first time I have seen in practice. Usually the problem is that observations are overdispersed with respect to the model.

Now, if data are underdispersed compared to a Poisson distribution (with equal mean and variance), I'd immediately think of a binomial distribution, where the variance $np(1-p)$ is always smaller than the mean $np$.

You could do a "kind of ad hoc binomial regression": parameterize $p=\frac{1}{1+e^{-\beta x}}$ with your regressors $x$ as in the logistic case, to constrain $0<p<1$. Set up the log-likelihood and estimate $\beta$. Do this separately for $n=1, 2, \dots$, yielding estimates $\hat{\beta}^{(n)}$. Pick the $(n, \hat{\beta}^{(n)})$ combination that yields the highest overall log-likelihood.

If your predictors are truly discrete and can only take on a finite number of values, you may instead want to group your data and estimate separate $(\hat{n}_x, \hat{\beta}_x)$ for each observed value of $x$ (then simply $\hat{p}_x = \frac{1}{1+e^{-\hat{\beta}_x}}$).

This is extremely ad hoc. Typically, in binomial regression, your $n$ is set a priori, and you are only interested in estimating your $p$ - here, you'd have to estimate both, and the problem is that there is no good interpretation of $\hat{n}$ (nor of $\hat{p}$). The only two good points this approach would have would be that (a) it's inherently discrete, in contrast to the normal approach that you mention, and (b) it should be able to capture the underdispersion. It still won't capture the negative correlation between innings (which you could potentially take into account by taking previous at-bats into your regressors), though. But it might improve on your model so far.

| cite | improve this answer | |
  • $\begingroup$ Hey Stephan, thanks so much. I don't usually deal with count data so I'm not very familiar with over/underdispersion. I'm fearful i'll run into the same problem with a binomial model but i'll give your method of picking n's and selecting the best model a shot. I believe the glm.nb function attempts to do just what you have described but i'll have to read up more on that. $\endgroup$ – Benjamin Whitesell Jun 2 '16 at 14:36
  • $\begingroup$ Good luck! And thank you for accepting - however, you may want to un-accept and see whether some better answer comes along (questions with accepted answers receive less traffic). Please feel free to do so. $\endgroup$ – Stephan Kolassa Jun 2 '16 at 14:42

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.