Censoring (clinical trials)

Linguistics

Engineering

Philosophy

Construct (python library)

Zero (linguistics)

Computer science

Mechanical engineering

Psychology

Mathematics

Political science

Social work

Work (physics)

Psychotherapist

Econometrics

Data science

Programming language

Tobit model

Meaning (existential)

Journal of the Society for Social Work and Research

Prevention researchers often analyze nonnegative continuous outcomes that are skewed and contain many zero observations. Variables with these features can arise when investigators measure low-base-rate behaviors (e.g., bullying, substance use) or use measurement tools that cannot differentiate individuals with low standing on the target construct. Common analytic procedures such as linear regression are often inappropriate for such data and may result in biased results. This article provides a tutorial on alternative methods that may be more suitable for the analysis of nonnormal, zero-bounded continuous outcomes. The meaning of the zero values is an important consideration when choosing an analytic approach. We discuss 2 common causes of abundant zeros (i.e., censoring due to measurement limits and true absence of the outcome), highlight problems that arise with the application of standard statistical methods to such outcomes, and describe promising but less well-known analytic alternatives. We illustrate these techniques using data from a study of bullying prevention in urban youth. Two approaches, Tobit regression and two-part models, are particularly useful methods for handling skewed nonnegative outcomes with several zero values. Familiarity with the issues and techniques we present may help researchers to make more informed analytic choices when confronted with such outcomes.

Analyzing Skewed Continuous Outcomes With Many Zeros: A Tutorial for Social Work and Youth Prevention Science Researchers