categorical: Class "categorical" and Inherited Classes

In mi: Missing Data Imputation and Model Checking

Class "categorical" and Inherited Classes

Description

The categorical class is a virtual class that inherits from the missing_variable-class and is the parent of the unordered-categorical and ordered-categorical classes. The ordered-categorical class is the parent of both the binary and interval classes. Aside from these facts, the rest of the documentation here is primarily directed toward developers.

Objects from the Classes

The categorical class is virtual, so no objects may be created from it. However, the missing_variable generic function can be used to instantiate an object that inherits from the categorical class by specifying type = "unordered-categorical", type = "ordered-categorical", type = "binary", type = "grouped-binary", or type = "interval".

Slots

The unordered-categorical class inherits from the categorical class and has no additional slots but must have more than two uniquely observed values in its raw_data slot. The default fit_model method is a wrapper for the multinom function in the nnet package. The ordered-categorical class inherits from the categorical class and has one additional slot:

cutpoints

Object of class "numeric" which is a vector of thresholds (sometimes estimated) that govern how an assumed latent variable is divided into observed ordered categories

The fit_model method for an ordered-categorical variable is, by default, a wrapper for bayespolr. The binary class inherits from the ordered-categorical class and has no additional slots. It must have exactly two uniquely observed values in its raw_data slot and its fit_model method is, by default, a wrapper for bayespolr. The grouped-binary class inherits from the binary class and has one additional slot:

strata

Object of class "character" which is a vector (possibly of length one) of variable names that group the observations into strata. The named external variables should also be categorical.

The default fit_model method for a grouped-binary variable is a wrapper for the clogit function in the survival package and the variables named in the strata slot are passed to the strata function.

The interval class inherits from the ordered-categorical class, has no additional slots, and is intended for variables whose observed values are only known up to orderable intervals. Its fit_model method is, by default, a wrapper for survreg even though it may or may not be a “survival” model in any meaningful sense.

Author(s)

Ben Goodrich and Jonathan Kropko, for this version, based on earlier versions written by Yu-Sung Su, Masanao Yajima, Maria Grazia Pittau, Jennifer Hill, and Andrew Gelman.

See Also

missing_variable

Examples

# STEP 0: GET DATA
data(nlsyV, package = "mi")

# STEP 0.5 CREATE A missing_variable (you never need to actually do this)
momrace <- missing_variable(as.factor(nlsyV$momrace), type = "unordered-categorical")
show(momrace)

mi documentation built on June 7, 2022, 1:04 a.m.