Algorithms to create, learn, fit and explore staged event tree models. Functions to compute probabilities, make predictions from the fitted models and to plot, analyze and manipulate staged event trees.
A staged event tree is a representation of a particular factorization of a joint probability over a product space. In particular, given a vector of categorical random variables \(X1, X2, \ldots\), a staged event tree represents the factorization \(P(X1, X2, X3, \ldots) = P(X1)P(X2 | X1) P(X3 | X1, X2) \ldots \). Additionally, the stages structure indicates which conditional probabilities are equal.
Model selection algorithms:
full model full
independence model indep
Hill-Climbing stages_hc
Backward Hill-Climbing stages_bhc
Fast Backward Hill-Climbing stages_fbhc
Backward Hill-Climbing Random stages_bhcr
Backward joining stages_bj
Hierarchical Clustering stages_hclust
K-Means Clustering stages_kmeans
Probabilities, log-likelihood and predictions:
Marginal probabilities prob
Log-Likelihood logLik.sevt
Predict method predict.sevt
Plot, explore and compare:
Plot plot.sevt
Compare compare_stages
Stages inclusion inclusions_stages
Stages info summary.sevt
Modify models:
Join and isolate unobserved situations join_unobserved
Join two stages join_stages
Rename a stage rename_stage
Collazo R. A., G<U+00F6>rgen C. and Smith J. Q. Chain event graphs. CRC Press, 2018.
G<U+00F6>rgen C., Bigatti A., Riccomagno E. and Smith J. Q. Discovery of statistical equivalence classes using computer algebra. International Journal of Approximate Reasoning, vol. 95, pp. 167-184, 2018.
Barclay L. M., Hutton J. L. and Smith J. Q. Refining a Bayesian network using a chain event graph. International Journal of Approximate Reasoning, vol. 54, pp. 1300-1309, 2013.
Smith J. Q. and Anderson P. E. Conditional independence and chain event graphs. Artificial Intelligence, vol. 172, pp. 42-68, 2008.
Thwaites P. A., Smith, J. Q. A new method for tackling asymmetric decision problems. International Journal of Approximate Reasoning, vol. 88, pp. 624<U+2013>639, 2017.
# NOT RUN {
data("PhDArticles")
mf <- full(PhDArticles, join_unobserved = TRUE)
mod <- stages_fbhc(mf)
plot(mod)
# }
Run the code above in your browser using DataLab