For the full paper, which includes all proofs, download the pdf here.
Categorical random variables are a common staple in machine learning methods and other applications across disciplines. Many times, correlation within categorical predictors exists, and has been noted to have an effect on various algorithm effectiveness, such as feature ranking and random forests. We present a mathematical construction of a sequence of identically distributed but dependent categorical random variables, and give a generalized multinomial distribution to model the probability of counts of such variables.
Bernoulli random variables are invaluable in statistical analysis of phenomena having binary outcomes, however, many other variables cannot be modeled by only two categories. Many topics in statistics and machine learning rely on categorical random variables, such as random forests and various clustering algorithms [6,7]. Many datasets exhibit correlation or dependency among predictors as well as within predictors, which can impact the model used. [6,9]. This can result in unreliable feature ranking , and inaccurate random forests .
Some attempts to remedy these effects involve Bayesian modeling  and various computational and simulation methods . In particular, simulation of correlated categorical variables has been discussed in the literature for some time [1, 3, 5]. Little research has been done to create mathematical framework of correlated or dependent categorical variables and the resulting distributions of functions of such variables.
Korzeniowski  studied dependent Bernoulli variables, formalizing the notion of identically distributed but dependent Bernoulli variables and deriving the distribution of the sum of such dependent variables, yielding a Generalized Binomial Distribution.
In this paper, we generalize the work of Korzeniowski  and formalize the notion of a sequence of identically distributed but dependent categorical random variables. We then derive a Generalized Multinomial Distribution for such variables and provide some properties of said distribution. We also give an algorithm to generate a sequence of correlated categorical random variables.