CONTACT US: Contact info
|
Citation |
Bangdiwala SI. Int. J. Inj. Control Safe. Promot. 2014; 21(3): 298-300. |
|
Affiliation |
Department of Biostatistics, UNC Gillings School of Global Public Health , Chapel Hill , NC , USA ; Institute for Social and Health Sciences , University of South Africa , Johannesburg , South Africa. |
|
Copyright |
(Copyright © 2014, Informa - Taylor and Francis Group) |
|
DOI |
|
PMID |
25078030 |
|
Abstract |
In the injury field, most variables of interest or count data (the number of pedestrian deaths in a given area in a given time period). There are some important variables, however, that are measured on a continuous scale, (blood-alcohol concentration) or constructed indices or scales that are treated as continuous variables (the injury severity score). Often, though, continuous variables are dichotomized for analyses. Dichotomy is a statistical procedure by which an originally continuous variable is transformed into a categorical binary variable based upon where individuals fall relative to a point. Dichotomization has been criticized on the grounds that at best it leads to a loss of a loss in analytic power and at worst, can create false or biased results. Unfortunately, many researchers continue to practice of dichotomous in their continuous variables prior to analysis. Arguments usually given include 1) it makes analysis simpler, 2) it makes presentation of results similar, 3) it eliminates the errors in measurement, and 4) it makes it easier to understand the resulting odds ratios or relative risks in epidemiological studies. The practice is quite common in epidemiology. Injury researchers may not be aware of the cost of dichotomization. Research has shown how dichotomization impacts the distributional properties of the continuous variable, which in turn affects the analytic method that can be used and thus, the interpretation of the results.... Language: en |