| 0406 Improving Correlation(s) Between Odor-judge Scores and Odor-gas Concentrations | ||
|
P.A. LENTON1, G.J. MAJERUS1, S. NACHNANI2, M. ROSENBERG3, and J.S. HODGES1, 1 University of Minnesota, School of Dentistry, Minneapolis, USA, 2 University Health Resources Group, Inc, Culver, CA, USA, 3 Tel Aviv University, Israel Oral Malodor assessment methods vary. On a 0-5 scale, some odor judges (OJs) use integers and some use fractions; it is uncertain if fractions improve the correlation between OJ scores and odor-gas concentrations. For objective measures of odor-gases using analytical devices, some investigators do and some do not log-transform the concentrations, obscuring comparisons between studies. Objective: Determine whether using fractions improves the correlation between OJ scores and concentrations, and whether OJ scores are more linearly related to concentration or log concentration. Methods: Six OJs rated eight compounds in varying concentrations, each compound was judged in its own session. Concentrations were presented in random order. OJs first gave an integer score, then a score with a fraction, if desired. For each gas and OJ, two correlations were computed: concentration vs. OJ scores, and common log concentration vs. OJ scores. Integer and fraction scoring were compared using four repeated-measures ANOVAs differing only in their dependent variables, with judges as subjects (random effect) and gas and scale (integer/fraction) as fixed effects. The four dependent variables were the two correlations above and their Fisher z-transforms. To compare concentration and log concentration, the same ANOVA was used with dependent variable {z-transform correlation of OJ score vs. log concentration} minus {z-transform correlation of OJ score vs. concentration}. Results: The correlation OJ scores vs. concentrations did not differ between integer and fraction scores (P > 0.25 for all four ANOVAs); this was consistent across gases and judges. OJ scores had higher correlations with log concentration than raw concentration (P<0.0001). Conclusion: Fractions did not improve the correlation between OJ scores and concentrations. Log concentration was more linearly related to OJ scores than was raw concentration. This study supported by the Centre for Research in Biomedicine, UWE, Bristol, UK and Veden Endowment for Clinical Research, UM, Mpls, MN USA.. | ||
| Seq #54 - Caries Detection/Periodontal Disease Assessment 11:00 AM-12:15 PM, Thursday, 26 June 2003 Svenska Massan Exhibition Hall B | ||
|
Back to the Diagnostic Systems Program
| ||