Conference Handouts

Click here to view handouts from 2010 - 2013 conferences. (Under agenda, then archives)

Click here to view past conference highlights.

Presented at the 21st Annual Research Conference

Handling Missing Data: The Motivation and Method of Multiple Imputation

Download Handouts: 11.5mb pdf

Presenting: Elizabeth Stuart

All Authors for this paper: Elizabeth Stuart; Melissa Azur; Constantine Frangakis; Philip Leaf

Presentation Type: element of symposium

Synopsis: Mental health services research is often hampered by missing data. Standard methods for dealing with missing data, such as complete-case analysis, rely on generally unreasonable assumptions about the missing values. This talk will discuss alternative approaches for handling missing data, focusing on multiple imputation (MI). MI is a flexible method that fully accounts for the uncertainty due to the missing values, and it is becoming increasingly easy to use.

Guidelines and Suggestions on How to Multiply Impute Missing Data

We're sorry, handouts are not available for this presentation.

Presenting: Melissa Azur

All Authors for this paper: Melissa Azur; Elizabeth Stuart; Constantine Frangakis; Philip Leaf

Presentation Type: element of symposium

Synopsis: Missing data is a common problem in mental health research. While a number of software packages have the capability of multiply imputing data, few resources are readily available to assist users in applying these methods. This presentation uses data from the Comprehensive Community Mental Health Services for Children and Their Families Program to discuss how multiple imputation techniques can be used in large data sets. Complications encountered and lessons learned are discussed.

Employing Multiply Imputed Data to Examine Disparities in Service Use Among Children

We're sorry, handouts are not available for this presentation.

Presenting: Crystal Barksdale

All Authors for this paper: Crystal Barksdale; Melissa Azur; Philip Leaf

Presentation Type: element of symposium

Synopsis: The issue of missing data has historically been a difficult and complicated issue in statistical analysis of mental health research. However, statistical methods have been developed to help researchers combat this very typical, yet frustrating, problem. The purpose of this study was to examine the association between race and past year mental health service use among children entering federally funded systems of care programs in the context of using a multiply imputed dataset.

Missing Data and Multiple Imputation: an Overview and Application of Techniques

We're sorry, handouts are not available for this presentation.

Session Number: 31 Room: Salon I

Presenting: Elizabeth Stuart

All Authors for this paper: Elizabeth Stuart

Presentation Type: brief symposium

Synopsis: Missing data occurs in nearly all children’s mental health services research. It is important to understand the consequences of missing data, as well as techniques for dealing with it. Traditional methods such as complete case analyses—which discard all individuals with any missing data—often rely on unreasonable assumptions and may lead to misleading results. This symposium describes a more principled approach for dealing with missing data—multiple imputation. The first talk in the symposium provides an overview of missing data, why it is a problem, and how to investigate missing data, as well as an overview of multiple imputation procedures. The second talk presents steps and suggestions on how to do multiple imputation, using freely available and easy to use software. The third talk discusses techniques for analyzing multiply imputed data and presents findings from a study that used multiply imputed data. The methods and ideas from each discussion are illustrated using data from the Children’s Mental Health Initiative, a federally funded program to develop systems of care for children and adolescents with serious emotional disturbances. All three talks focus on implications and practical applications of these methods in order to make them more accessible to a wide range of researchers.

Date:

Session Time: 3:45 PM - 4:45 PM