Abstract
We propose a simplified version of the partially observed quasi-information matrix (Poquim) method for inference about non-Gaussian linear mixed models and show its computational advantage over the original method. We illustrate the difference, and compare performance of the simplified version with Poquim as well as the normality-based method in simulation studies. An example of real-data analysis is considered.
Original language | English (US) |
---|---|
Pages (from-to) | 171-189 |
Number of pages | 19 |
Journal | Computational Statistics |
Volume | 38 |
Issue number | 1 |
DOIs | |
State | Published - Mar 2023 |
Keywords
- Asymptotic covariance matrix
- Fisher information
- Non-Gaussian linear mixed model
- Quasi-likelihood
- REML
- Spoquim
ASJC Scopus subject areas
- Statistics and Probability
- Statistics, Probability and Uncertainty
- Computational Mathematics