Tobit model for outcome variable is limited by censoring in nursing research

In nursing research, censoring may limit the outcome variable of interest. Well-known examples reflect the so-called floor or ceiling effect. Offering a biased estimator of the limited data, the linear regression model with ordinary least-squares estimation is one of the most frequently used methods. This method requires an innovative approach.

The aims of this study were to (a) introduce an alternative data analysis approach (Tobit model), (b) discuss the constraints of ordinary least-squares regression and complementary advantages of the Tobit model in working with censoring data, and (c) provide a fundamental understanding of the Tobit model in the nursing and healthcare domains.

The benefits of the Tobit model are outlined using an example of activities of daily living.

Based on model-fit parameters, residual plots, regression coefficients, and significance levels, the results indicate that the Tobit model performs more accurately than an ordinary least-squares regression when data are subject to a ceiling or floor effect.

When the Tobit model is specified correctly, it allows nursing researchers to ameliorate biased coefficient estimates occurring when health status measures are subject to limited data by censoring. Despite the theoretical superiority of the Tobit model to classical statistical data analysis, nurse researchers are cautioned to be aware of the limitations and appropriate applications associated with the Tobit model.