Builds the woe dictionary of a set of predictor variables upon a given binary outcome. Convenient to make a woe version of the given set of predictor variables and also to allow one to tweak some woe values by hand.
dictionary(.data, outcome, ..., Laplace = 1e-06)
A tbl. The data.frame where the variables come from.
The bare name of the outcome variable with exactly 2 distinct values.
bare names of predictor variables or selectors accepted by dplyr::select()
.
Default to 1e-6. The pseudocount
parameter of the Laplace Smoothing
estimator. Value to avoid -Inf/Inf from predictor category with only one outcome class.
Set to 0 to allow Inf/-Inf.
a tibble with summaries and woe for every given predictor variable stacked up.
You can pass a custom dictionary to step_woe()
. It must have the exactly
the same structure of the output of dictionary()
. One easy way to do this
is by tweaking an output returned from it.
Kullback, S. (1959). Information Theory and Statistics. Wiley, New York.
Hastie, T., Tibshirani, R. and Friedman, J. (1986). Elements of Statistical Learning, Second Edition, Springer, 2009.
Good, I. J. (1985), "Weight of evidence: A brief survey", Bayesian Statistics, 2, pp.249-270.
# NOT RUN {
mtcars %>% dictionary(am, cyl, gear:carb)
# }
Run the code above in your browser using DataLab