| Artificial_3 {SparseMSE} | R Documentation |
Artificial data set to demonstrate possible instabilities
Description
This is a simple data set based on three lists, which gives examples of models that
fail on one or the other of the criteria tested by checkident. This is Table 2 in Chan, Silverman and Vincent (2019).
Usage
Artificial_3
Format
An object of class data.frame with 4 rows and 4 columns.
Details
If all three two-list effects are included in the fitted model then the linear program
in checkident yields a strictly positive value but the matrix A is not of full column rank, so the parameters are not identifiable.
If the model contains AB either alone or in conjunction with one of AC and BC, then the linear program result is zero, so the MLE does not exist.
If only main effects are considered, or if either or both of AC and BC, but not AB are included,
then the model passes both tests.
References
Chan, L., Silverman, B. W., and Vincent, K. (2019). Multiple Systems Estimation for Sparse Capture Data: Inferential Challenges when there are Non-Overlapping Lists. Available from https://arxiv.org/abs/1902.05156.