Parsimonious Mixed Models

Published on Jun 16, 2015in arXiv: Methodology
Douglas M. Bates32
Estimated H-index: 32
Reinhold Kliegl59
Estimated H-index: 59
+ 1 AuthorsHarald Baayen7
Estimated H-index: 7
(University of Tübingen)
View in Source
The analysis of experimental data with mixed-effects models requires decisions about the specification of the appropriate random-effects structure. Recently, Barr, Levy, Scheepers, and Tily, 2013 recommended fitting `maximal' models with all possible random effect components included. Estimation of maximal models, however, may not converge. We show that failure to converge typically is not due to a suboptimal estimation algorithm, but is a consequence of attempting to fit a model that is too complex to be properly supported by the data, irrespective of whether estimation is based on maximum likelihood or on Bayesian hierarchical modeling with uninformative or weakly informative priors. Importantly, even under convergence, overparameterization may lead to uninterpretable models. We provide diagnostic tools for detecting overparameterization and guiding model simplification.
Figures & Tables
  • References (19)
  • Citations (188)
#1Douglas M. BatesH-Index: 32
#2Martin MächlerH-Index: 16
Last.Steve WalkerH-Index: 2
view all 4 authors...
14.9k CitationsSource
#1Jacob Westfall (CU: University of Colorado Boulder)H-Index: 9
#2Charles M. Judd (CU: University of Colorado Boulder)H-Index: 61
Last.David A. Kenny (UConn: University of Connecticut)H-Index: 65
view all 3 authors...
39 CitationsSource
#1Simon N. Wood (University of Bath)H-Index: 42
120 CitationsSource
#1Andries F. SandersH-Index: 1
#2Andries SandersH-Index: 1
259 CitationsSource
1,997 CitationsSource
#1R. Harald Baayen (U of A: University of Alberta)H-Index: 45
#2Petar Milin (University of Novi Sad)H-Index: 11
325 CitationsSource
#1R. Harald Baayen (U of A: University of Alberta)H-Index: 45
1,656 Citations
#1Edmundo Kronmüller (UCR: University of California, Riverside)H-Index: 5
#2Dale J. Barr (UCR: University of California, Riverside)H-Index: 24
66 CitationsSource
#1L. de Vaan (Radboud University Nijmegen)H-Index: 1
#2Robert SchreuderH-Index: 43
Last.R.H. BaayenH-Index: 4
view all 3 authors...
45 CitationsSource
#1Shravan Vasishth (University of Potsdam)H-Index: 23
#2Richard L. LewisH-Index: 31
131 CitationsSource
Cited By188
#1Lea-Sophie Borgmann (RKI: Robert Koch Institute)H-Index: 1
#2Lars Eric KrollH-Index: 19
Last.Thomas Lampert (RKI: Robert Koch Institute)H-Index: 38
view all 5 authors...
#1Romy Frömer (Brown University)H-Index: 5
#2Carolyn K. Dean Wolf (Brown University)H-Index: 1
Last.Amitai Shenhav (Brown University)H-Index: 14
view all 3 authors...
2 CitationsSource
#1My-Lan T. Le (California Polytechnic State University)H-Index: 1
#2Christopher M. Garvin (California Polytechnic State University)
Last.Clinton D. Francis (California Polytechnic State University)H-Index: 17
view all 4 authors...
#1Ariana Orvell (UM: University of Michigan)H-Index: 4
#2Ethan Kross (UM: University of Michigan)H-Index: 35
Last.Susan A. Gelman (UM: University of Michigan)H-Index: 65
view all 3 authors...
#1Francesco Mantegna (NYU: New York University)
#2Florian Hintz (MPG: Max Planck Society)H-Index: 4
Last.Falk Huettig (Radboud University Nijmegen)H-Index: 24
view all 5 authors...
#1E. A. Issaly (National University of Cordoba)
#2Alicia N. Sérsic (National University of Cordoba)H-Index: 18
Last.Valeria Paiaro (National University of Cordoba)H-Index: 5
view all 7 authors...
#1Vasilis Kokkoris (UBC: University of British Columbia)H-Index: 3
#2Miranda M. Hart (UBC: University of British Columbia)H-Index: 35
1 CitationsSource
#1Jarkko HautalaH-Index: 4
#2Otto Loberg (University of Jyväskylä)H-Index: 2
Last.Paavo H. T. Leppänen (University of Jyväskylä)H-Index: 35
view all 6 authors...
View next paperFitting Linear Mixed-Effects Models Using lme4