Division of Psychiatric Statistics (DPS)

The goal of the Division of Psychiatric Statistics (DPS) is to provide and promote innovative biostatistics collaborations and data coordinating services in support of research on health and health-related behavior outcomes. Over 160 papers in peer-reviewed journals, 7 book chapters and 3 books have been published by DPS personnel over the past 5 years. (see “Publications” for details). DPS has also helped with over 100 grant submissions with investigators in Psychiatry and other departments within the UR Medical Center and numerous projects during the same period. The key DPS personnel all have many years of experience (ranging from 5 to 20 years) in grant preparation (particular for NIH grants), trial execution and project analysis. 

To overcome barriers of interdisciplinary research, we

  1. Provide consultation on study design, power analysis, statistical analyses, interpretation of results, and grant and manuscript preparation;
  2. Supplement investigators’ awareness of modern and cutting-edge statistical methods and facilitate exchange among investigators concerning the design and analysis implications of such methods;
  3. Perform integrated data preparation and data analysis using software packages such as MPlus, R, SAS, STATA, Winbugs, MULTILOG and LISREL, as well as the programs we have developed to facilitate research for a range of topics in psychosocial, and integrated health and behavior research, including but not limited to:

Modeling Dyadic and Interdependent Data

 

Survival Analysis

 

Equivalence and Non-inferiority Testing

 

Effect Size and Power Analysis

 

Social Network Analysis

Models for Instrumentation

  • Principal Component Analysis and Factor Analysis (exploratory or confirmatory) for scale construction
  • ROC Analysis for diagnostic and screening tests
  • Models for internal item and scale-level consistency such as Cronbach coefficient alpha and intra-class correlation
  • Agreement and Reliability analysis such as Cohen’s kappa, Concordance Correlation Coefficient, and ICC.

Structural Equation Models (SEM)

  • Mediation Analysis
  • General structural relationships involving multiple endogenous and exogenous variables.

Models for Mixture Study Populations

  • Zero-modified models (ZIP) for count data including zero-inflated Poisson (ZIP), zero-inflated negative binomial (ZINB), Hurdle Model,
  • Cluster analysis
  • Growth Mixture Models (GMM) for Moderation Analysis

Models for Longitudinal Data Analysis

  • Generalized Linear Mixed-effects Models (GLMM) for multi-level studies
  • Weighted Generalized Estimating Equations (WGEE) for robust inference with skewed data

Models for Causal Inference

  • Methods for addressing selection bias in non- and semi-randomized studies such as Propensity Score and Marginal Structural Models (MSM). 
  • Methods for addressing confounders in post-randomization for randomized controlled trials (RCT) such as and Principal Stratification methods and Structural Nested Models (SNM)
  • Doubly Robust estimation of treatment effects
  • Instrumental Variable (IV) method for addressing selection bias
  • Cost and Cost-effectiveness analysis using Propensity Score

Latent class analysis

  • Latent class analysis for continuous and categorical outcomes
  • Latent class scaling analysis
  • Latent transition analysis

Modeling Dyadic and Interdependent Data

  • Dyadic data for continuous and categorical outcomes

Survival Analysis

  • Parametric and Semi-parametric (Cox) regression for continuous and discrete outcomes
  • Repeated Events (Multiple Event Time data)
  • Cure Models for population mixtures

Equivalence and Non-inferiority Testing

  • Parametric and non-parametric tests for equivalence
  • Methods for continuous and non-continuous outcomes
  • Confidence-interval inclusion rules for equivalence
  • Equivalence tests for paired observations

Effect Size and Power Analysis

  • Power analysis for mediation analysis
  • Power analysis for longitudinal data
  • Power analysis for stepped wedge study design
  • Within- and between-group effect size estimates for cross section and longitudinal data

Social Network Analysis

  • Computations for social network metrics (e.g., density, centrality and Pagerank)
  • Agent based models
  • Social network diffusion models
  • Models for social network connectivity

top