validate {igate}

Validates results after using igate or categorical.igate.

Description

Takes a new data frame to be used for validation and the causes and control bands obtained from igate or categorical.igate and returns all those observations that fall within these control bands.

Usage

validate(validation_df, target, causes, results_df, type = NULL)

Arguments

validation_df

Data frame to be used for validation. It is recommended to use a different data frame from the one used in igate/ categorical.igate. The same data frame can be used if just a sanity check of the results is performed. This data frame must contain the target variable as well as all the causes determined by igate/ categorical.igate.

target

Target variable that was used in igate or categorical.igate.

causes

Causes determined by igate or categorical.igate. If you saved the results of igate/ categorical.igate in an object results, simply use results$Causes here.

results_df

The data frame containing the results of igate or categorical.igate.

type

The type of igate that was performed: either "continuous" or "categorical". If not provided function will try to guess the correct type based on the type of validation_df[[target]].

Details

If a value of Good_Count or Bad_count is very low in the second data frame, it means that this cause is excluding a lot of observations from the first data frame. Consider re-running validate with this cause removed from causes.

Value

A list of three data frames is returned. The first data frame contains those observations in validation_df that fall into *all* the good resp. bad control bands specified in results_df. The columns are target, then one column for each of the causes and a new column expected_quality which is "good" if the observation falls into all the good control bands and "bad" if it falls into all the bad control bands.

The second data frame has three columns

Cause	Each of the causes.
Good_Count	If we selected all those observations that fall into the good band of this cause, how many observations would we select?
Bad_Count	If we selected all those observations that fall into the bad band of this cause, how many observations would we select?

The third data frame summarizes the first data frame: If type = "continuous" it has three columns:

expected_quality	Either "good" or "bad".
max_target	The maximum value for target for the observations with "good" expected quality resp. "bad" expected quality.
min_target	Minimum value of target for good resp. bad expected quality.

If type = "categorical" it has the following three columns:

expected_quality	Either "good" or "bad".
Category	A list of categories of the observations with expected quality good resp. bad.
Frequency	A count how often the respective Category appears amongs the observations with good/ bad expected quality.

Examples

validate(iris, target = "Sepal.Length", causes = resultsIris$Causes, results_df = resultsIris)