Mixture model for analysis of tuberculin surveys
Savanur, Shashidhar J.; Chadha, V.K. and Jagannatha, P.S. (2002) Mixture model for analysis of tuberculin surveys. Indian Journal of Tuberculosis, 49. pp. 147-152.
Full text available as:
Setting : Data obtained in four tuberculin surveys by National Tuberculosis Institute, Bangalore. Objectives : 1. To estimate the prevalence of tuberculosis infection by using a mixture model for analysis. 2. To compare the estimates of relevance of infection obtained by the mixture model analysis and computed ARI with the estimates based on conventional methods. Design : The tuberculin reaction sizes in a given distribution were assumed to represent a mixture of three sub-groups, each following a normal distribution. The maximum likelihood estimates of the parameters of component distributions and the proportion of the sub-groups in the study population were obtained using EM-algorithm. The estimates of prevalence of tuberculosis infection obtained by mixture model analysis were compared with the estimates based on locating the cut-off point (anti-mode) or the mode of reactions attributable to infection with M.tuberculosis. Results :The prevalence of infection estimates obtained by mixture model for different age groups in the four surveys were similar to the estimates obtained based on the anti-mode or the mode method. Conclusion :The mixture model analysis can be a very useful tool for analysis of tuberculin surveys, specially when there is difficulty in identification of the anti-mode or the mode of reactions attributable to infection with M. tuberculosis.
Archive Staff Only: edit this record