Conditional and unconditional prediction for censored ordered probit

Share:

Description

Conditional and unconditional prediction for censored ordered probit. Unconditional prediction returns the fitted values (predicted probabilities) from the cpolr object. Conditional prediction takes the observed range of the diff-corrected self-response output from anchors and renormalizes the predicted probabilities for each observation.

Usage

1
2
   ## S3 method for class 'cpolr'
fitted(object, anchors, average = FALSE, unconditional = FALSE, ...)

Arguments

object

output from cpolr.

anchors

leave missing for unconditional prediction (or set unconditional=TRUE). For conditional prediction, specify the object of class anchors.rank used to run cpolr originally.

average

a logical value. See values below for more details.

unconditional

Set to TRUE if you submit an anchors.object AND want the unconditional probabilities returned. One case that you would submit a anchors.rank object is if you did subsetting for the anchors object but not for the cpolr object, and want the intersection of the two objects used for the unconditional probabilities.

...

required for S3, but any other options will be ignored.

Value

If average = FALSE, a matrix of predicted probabilities with rows corresponding to observations, and columns corresponding to categories.

If average = TRUE, the matrix of predicted probabilities (conditional or unconditional) is summarized to a vector (summed by categories, then renormalized to sum to 1).

If anchors object has been specified, then each observation is renormalized to fall into the range of the diff-corrected self-response for that observation. If there are no ties for a given observation, then that observation is a vector consisting of (k-1) zeros and 1 one. If there are ties, then the predicted probabilities for that observation are renormalized to fall within the diff-corrected range.

If anchors object is omitted, identical to the matrix of predicted probabilities from the cpolr output.

Note

If the anchors object is made using a subset of the data used to create the cpolr object, then invoking fitted.cpolr will not use the same cases in calculating the conditional probabilities as it would if the anchors object is omitted!

If you want to have the same cases used in the unconditional calculation as in the conditional with a subsetted anchors object, then include anchors object and set unconditional.override = TRUE.

Related materials and worked examples are available at http://wand.stanford.edu/anchors/

Author(s)

Jonathan Wand http://wand.stanford.edu

References

Venables, W. N. and Ripley, B. D. (2002) Modern Applied Statistics with S. 4th edition. Springer.

Wand, Jonathan; Gary King; and Olivia Lau. (2007) “Anchors: Software for Anchoring Vignettes”. Journal of Statistical Software. Forthcoming. copy at http://wand.stanford.edu/research/anchors-jss.pdf

Wand, Jonathan and Gary King. (2007) Anchoring Vignetttes in R: A (different kind of) Vignette copy at http://wand.stanford.edu/anchors/doc/anchors.pdf

Gary King and Jonathan Wand. "Comparing Incomparable Survey Responses: New Tools for Anchoring Vignettes," Political Analysis, 15, 1 (Winter, 2007): Pp. 46-66, copy at http://gking.harvard.edu/files/abs/c-abs.shtml.

See Also

anchors, cpolr

Examples

1
## Basic usage: see cpolr

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.