Premium
A note on the use of the Tobit approach for tests scores with floor or ceiling effects
Author(s) -
Oord Edwin J. C. G.,
Ark L. Andries
Publication year - 1997
Publication title -
british journal of mathematical and statistical psychology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.157
H-Index - 51
eISSN - 2044-8317
pISSN - 0007-1102
DOI - 10.1111/j.2044-8317.1997.tb01150.x
Subject(s) - tobit model , censoring (clinical trials) , statistics , econometrics , mathematics , likert scale
Muthén (1989) developed the Tobit approach to deal with censored variables in structural equation modelling. We performed a simulation study to evaluate the effectiveness of this approach for the type of censoring that occurs for sums of Likert items. Results indicated that estimated correlations and factor loadings were substantially biased downwards, and that goodness‐of‐fit tests were inaccurate. Furthermore, the magnitude of distortion depended on characteristics of the scales such as the amount of censoring, scale length, the number of response categories of the items, and the size of the correlation. When we adjusted the Tobit approach by assuming that all scores at the lower end of the scale were affected by censoring and not just the lowest score, results improved considerably. In addition, this adjusted Tobit approach was superior to alternative procedures that deal with censoring and was easy to implement using the existing software.