-
Tytuł:
-
Mixed-type multivariate response regression with covariance estimation.
-
Autorzy:
-
Ekvall KO; Division of Biostatistics, Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden.; Applied Statistics Research Unit, Institute of Statistics and Mathematical Methods in Economics, TU Wien, Vienna, Austria.
Molstad AJ; Department of Statistics and Genetics Institute, University of Florida, Gainesville, Florida, USA.
-
Źródło:
-
Statistics in medicine [Stat Med] 2022 Jul 10; Vol. 41 (15), pp. 2768-2785. Date of Electronic Publication: 2022 Mar 24.
-
Typ publikacji:
-
Journal Article; Research Support, Non-U.S. Gov't; Research Support, U.S. Gov't, Non-P.H.S.
-
Język:
-
English
-
Imprint Name(s):
-
Original Publication: Chichester ; New York : Wiley, c1982-
-
MeSH Terms:
-
Algorithms*
Research Design*
Humans ; Linear Models
-
References:
-
Hum Reprod Update. 2014 Jan-Feb;20(1):124-40. (PMID: 24077980)
Stat Med. 2008 Sep 30;27(22):4408-27. (PMID: 18551509)
Stat Med. 2006 Apr 30;25(8):1307-22. (PMID: 16217846)
Stat Med. 1997 Apr 30;16(8):883-900. (PMID: 9160486)
Biometrics. 1996 Jun;52(2):740-50. (PMID: 8672710)
Stat Med. 2007 Sep 10;26(20):3782-800. (PMID: 17133630)
Stat Med. 2022 Jul 10;41(15):2768-2785. (PMID: 35699353)
Biometrics. 1997 Mar;53(1):110-22. (PMID: 9147588)
-
Grant Information:
-
P 30690 Austria FWF_ Austrian Science Fund FWF; P30690-N35 Austria FWF_ Austrian Science Fund FWF
-
Contributed Indexing:
-
Keywords: covariance estimation; latent variable models; mixed-type response regression; multivariate regression
-
Entry Date(s):
-
Date Created: 20220614 Date Completed: 20220621 Latest Revision: 20230928
-
Update Code:
-
20240105
-
PubMed Central ID:
-
PMC9313904
-
DOI:
-
10.1002/sim.9383
-
PMID:
-
35699353
-
We propose a new method for multivariate response regression and covariance estimation when elements of the response vector are of mixed types, for example some continuous and some discrete. Our method is based on a model which assumes the observable mixed-type response vector is connected to a latent multivariate normal response linear regression through a link function. We explore the properties of this model and show its parameters are identifiable under reasonable conditions. We impose no parametric restrictions on the covariance of the latent normal other than positive definiteness, thereby avoiding assumptions about unobservable variables which can be difficult to verify in practice. To accommodate this generality, we propose a novel algorithm for approximate maximum likelihood estimation that works "off-the-shelf" with many different combinations of response types, and which scales well in the dimension of the response vector. Our method typically gives better predictions and parameter estimates than fitting separate models for the different response types and allows for approximate likelihood ratio testing of relevant hypotheses such as independence of responses. The usefulness of the proposed method is illustrated in simulations; and one biomedical and one genomic data example.
(© 2022 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.)