Joshua Wiley is a lecturer in the Monash Institute for Cognitive and Clinical Neurosciences and School of Psychological Sciences at Monash University. He completed my PhD at the University of California at Los Angeles in Psychology with a major in health psychology research, and did a post doc at the Mary MacKillop Institute for Health Research in primary care and prevention research. He studies the roles of risk and protective factors in the psychological and biological response to stress and chronic disease. Specifically, his interests are in affective science and understanding the protective role of emotion regulation factors including psychosocial resources (e.g., optimism, social support) and approach and avoidance-oriented coping processes. He studies whether and how these factors promote adjustment and health to stressful experiences including: cancer, low socioeconomic status, and general life stress. Examples of questions his research addresses are:
I am fascinated by the individual differences in how people react to and cope with stressful events. This has led me to explore the roles of psychosocial resources: psychological and social factors that are innately valuable or that individuals can draw on to promote good mental or physical health. Specific resources I have studied include social support, self-efficacy, and positive outcome expectancies. Given the degree of measurement overlap of many resources, another line of my research has been examining measurement and factor analytic models of constellations of psychosocial resources. In four samples analyzed so far, a two-factor model with a latent psychological and social resource factor has fit well.
Coping strategies people employ in response to stressors also play a role in how well they adjust. Coping is particularly important in the face of major life stressors such as cancer or chronic disease. In collaboration with the UCLA Stanton Stress and Coping lab, I have studied the effects of coping in response to genetic testing, breast cancer diagnosis, and an experimental study of receipt of negative health information.
To understand how psychological and physiological processes interact, I have worked on studies measuring biomarkers such as salivary cortisol and alpha-amylase to assess HPA axis and sympathetic nervous system reactivity to stress, and whether psychosocial resources predict physiological baseline or reactivity to acute stress. I collaborate with researchers at Monash University, Stanford, and the University of Melbourne processes related to sleep, including the roles of cognitive vulnerabilities in self- and actigraphy-assessed sleep, and the effects of variability in sleep.
As the number of biomarkers available has grown, researchers are increasingly faced with how to choose which to collect or if many are collected how to model them. Given my work exploring measurement models of psychosocial resources, I am interested in applying similar approaches to biomarkers. I am particularly interested in practical analytical models researchers can use not only to reduce the number of dimensions of biomarkers for analysis, but also to find theoretically meaningful dimensions. Related questions include how best to utilize repeated measures of biomarkers, issues of non-normality or non-linearity, and adjusting for potential biasing effects of medications and other measurement error.
I actively collaborate on interdisciplinary teams. In the area of health services, I have engaged in research probing the frequency and cost of providing futile care in critical care units. We also investigated predictors of providing care assessed as futile: what characteristics of patients or providers increases the probability that a patient will receive futile care? Do different types of providers such as attending physicians, fellows, and nurses ahve different perceptions of when or for which patients, care is futile?
A separate line of work has involved modeling and understanding psychiatric conditions in children and adolescents. Projects have included scale design and assessment of anxiety in children as well as analysis of behavioral and EEG patterns that are signatures of ADHD. In the ADHD research I am involved in, we have tried to characterize differences in networks between children with and without ADHD, rather than only looking at differences on single measures. Most recently, I am involved in a project pursuing novel, interactive, behavioral measures of ADHD.
As senior partner at Elkhart Group Ltd., Joshua advises on data analytics for research in academia and industry. Through Elkhart Group Ltd. and former work at UCLA Statistical Consulting Group, he has experience with a wide array of clients ranging from graduate students, to experienced researchers, and biotechnology companies.
Deriving unbiased, reliable measures is a common problem
in psychology where core constructs such as depression and
personality are unobservable. I integrate psychometrics and
item response theory into my work to create latent variables
of focal constructs and to explore their accuracy and
reliability of these measures. I have used confirmatory
factor models to understand the latent factor structure
underlying psychosocial resources (Wiley, ..., Stanton,
under review) and bi-factor models to separate variability
associated with anxiety from multiple reporters (parent and
child) in categorical response data. Although most of work
employs maximum likelihood methods, I am also interested in
Bayesian methods. In one study, I used Bayesian factor
analysis to extract multiple draws of factor scores from the
posterior distribution of the latent variable to
meta-analyze several samples where the latent construct was
assumed similar, but the specific indicators varied (Wiley,
..., Stanton, under review). I am also exploring the use
of Bayesian factor analysis with strong priors to allow
small covariances among residuals. I use a variety of
analytical software including the commercial
Mplus as well
as open source variants such as
Bayesian models, I have used both
Mplus and coded my own
factor models in
have also written a small number of utility functions
related to measurement models such as getting maximum
likelihood estimates of descriptive statistics using all
available data in the
semutils available on
github, and I am a co-developer of the
MplusAutomation to link
also available on
has been cited by over 15 publications (see
and is in the top 15% of most
Another focus of my statistics work is on mixed and
random effects models (e.g., patients sampled within
physicians; longitudinal observational studies). One
project I worked on involved observations crossed between
patients (repeated measures) and physicians and the outcome
was ordinal (Huynh, Kleerup, Wiley, ..., Wenger, 2013).
I analyzed the data in a Bayesian framework and presented
the average marginal change in the probability of falling
into each level of the outcome as an intuitive way for
clinicians to understand the results and effect of each
variable. To automate generating average marginal
predictions following an ordinal, cross-classified Bayesian
mixed effects model, I wrote the
available on github and presented a poster on the work at
Modern Modeling Methods 2013 meeting.
I have expertise using multiple imputation and maximum likelihood approaches to more efficiently use data in the case they are missing completely at random (MCAR) or at least missing at random (MAR). However, in longitudinal studies, attrition is ubiquitous and assuming this is random is often untenable. I have done work using joint models or so-called shared parameter models where the random intercept and slope parameters from a growth model are used as predictors of the probability of still being in the study (Wiley, ..., Stanton, in preparation). I have also used joint models in work with clients where the outcome was one of the decision criteria for whether participants could remain in the study (i.e., non random as participants who worsened too quickly were triaged out of the study).
I am currently working on developing methods for
analyzing and modeling intra-individual variability. I have
conducting variability analyis.
Currently I am working on extending the software to allow
more flexible models, alternative distributions, and to use
variability as both a predictor and as an outcome (or both,
as in mediation).
In collaboration with other researchers, we are also working
on publishing a number of empirical papers exploring the
effects of variability.
Areas of current interest include models and tests for
multilevel mediation. In my work and consultation, I have
used a non-parametric two-stage re-sampling bootstrap
approach as well as a fully Bayesian model derive confidence
intervals around the indirect effect. I have conducted some
small simulations examining the properties of two-stage
re-sampling for multilevel models and code to make use of
multiple processors or clusters to speed non parametric
bootstrapping of multilevel models
to use, fast, tools for applied research using multilevel
mediation models is an active area of interest. I
am also interested in techniques to estimate the
distribution of indirect effects when one of the pathways is
a random effect and in how this information can be presented
to be useful to researchers, rather than marginalizing out
To get in touch with me about my research or a research collaboration please contact me here. If you have questions about statistical analysis or would like a statistical analyst or consultant on your project, please go to Elkhart Group Ltd.
Copyright © 2009 - 2016 by Joshua Wiley. All rights reserved. Brought to you by Elkhart Group Limited.