Extended information criterion (EIC) approach for linear mixed effects models under restricted maximum likelihood (REML) estimation.

Akifumi Yafune; Takashi Funatogawa; Makio Ishiguro

doi:10.1002/sim.2191

Home/Archive/PubMed

PubMed · 16237658

Extended information criterion (EIC) approach for linear mixed effects models under restricted maximum likelihood (REML) estimation.

Akifumi Yafune·Takashi Funatogawa·Makio Ishiguro

View original source ↗Export metadata ↓

Abstract

In clinical data analysis, the restricted maximum likelihood (REML) method has been commonly used for estimating variance components in the linear mixed effects model. Under the REML estimation, however, it is not straightforward to compare several linear mixed effects models with different mean and covariance structures. In particular, few approaches have been proposed for the comparison of linear mixed effects models with different mean structures under the REML estimation. We propose an approach using extended information criterion (EIC), which is a bootstrap-based extension of AIC, for comparing linear mixed effects models with different mean and covariance structures under the REML estimation. We present simulation studies and applications to two actual clinical data sets.

Explore related subjects

Biometry Growth Humans Likelihood Functions Linear Models Longitudinal Studies Platelet Count Purpura, Thrombocytopenic, Idiopathic Journal Article Research Support, Non-U.S. Gov't

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Akifumi Yafune, Takashi Funatogawa, Makio Ishiguro. 2005-11-30. Extended information criterion (EIC) approach for linear mixed effects models under restricted maximum likelihood (REML) estimation.. https://doi.org/10.1002/sim.2191

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2007

The role of education in biostatistical consulting.

Medical students, residents, postdoctoral fellows, and faculty commonly consult with biostatistical experts about study design and data analysis when conducting clinical research. The role of biostatistical training during these consultations is examined, and characterizations of the connections between biostatistical consultation and education are reviewed. The presence and kinds of teaching efforts during biostatistical consults at four academic research institutions over various periods of time between 1999 and 2005 (237 consultations in total) were recorded and are described. By site, 67, 70, 78, and 100 per cent of the consulting sessions included biostatistical training, with an overall 78 per cent (95 per cent CI: 73-83 per cent) of consultations including an educational component when all consultations were combined. Training covered a wide range of biostatistical topics. Seventy-five per cent of the consultations with faculty (120/161), 79 per cent with fellows and residents (31/39), and 100 per cent with medical students (10/10) included some degree of instruction in study design or statistical analysis topics. Results show that both the need and the opportunity exist for specialized biostatistical instruction during one-on-one sessions between a consulting biostatistician and physicians, medical students, and research staff. Academic researchers are ideally positioned to absorb this kind of training when they initiate a request for assistance with their own research project.

Biometry↗

PubMed2007

Improving the quality of patient care using reliability measures: a classification tree approach.

This paper considers the application and interpretation of new reliability measures for a classification tree-based medical risk assessment tool. Following the construction of a classification tree reliability measures may then be used to provide an estimate of the precision of the classification and the probability in each terminal node of the classification tree. Identification of unreliable nodes (those that have low precision) in this application may indicate patient groups requiring closer monitoring or scenarios in which further information about the patient is required, thereby providing medical practitioners with an avenue for more informed decision making.

Biometry↗

PubMed2007

Confidence intervals for the amount of heterogeneity in meta-analysis.

Effect size estimates to be combined in a systematic review are often found to be more variable than one would expect based on sampling differences alone. This is usually interpreted as evidence that the effect sizes are heterogeneous. A random-effects model is then often used to account for the heterogeneity in the effect sizes. A novel method for constructing confidence intervals for the amount of heterogeneity in the effect sizes is proposed that guarantees nominal coverage probabilities even in small samples when model assumptions are satisfied. A variety of existing approaches for constructing such confidence intervals are summarized and the various methods are applied to an example to illustrate their use. A simulation study reveals that the newly proposed method yields the most accurate coverage probabilities under conditions more analogous to practice, where assumptions about normally distributed effect size estimates and known sampling variances only hold asymptotically.

Biometry↗