Assessment of digit preference in self-reported year at menopause: choice of an appropriate reference distribution
Estimated associations between self-reported age at menopause and health may be attenuated if self-report is inaccurate. In a cross-sectional study, it is not possible to assess validity or reproducibility. Instead, one can examine digit preference, e.g., overreporting of numbers ending in zero or five. Typically, analyses use equal proportions-10% probability for each digit-as the reference distribution. Depending on the age distribution of the sample and on the underlying distribution of age at the event, however, an alternative reference distribution may be more appropriate. As an illustration, the authors examined digit preference in the self-reported calendar year at the final menstrual period in cross-sectional data from 2,151 naturally postmenopausal women in the Study of Women's Health Across the Nation (1995-1997), a multisite, multiethnic study of women aged 40-55 years. With chi-square tests, the distribution of terminal digit for self-reported year at the final menstrual period was compared with several reference distributions. The observed distribution was much closer to a reference distribution based on previously published data than it was to equal proportions. Future assessments of digit preference in cross-sectional studies should consider alternatives to equal proportions, particularly for samples with small age ranges and events with a restricted underlying age distribution.
Sybil L. Crawford, Catherine B. Johannes, and Rebecca K Stellato. "Assessment of digit preference in self-reported year at menopause: choice of an appropriate reference distribution" American journal of epidemiology 156.7 (2002).
Available at: http://works.bepress.com/crawfords/2