Analisis Data Longitudinal dengan Metode Regresi Berstruktur Pohon (Kasus Penyakit Kencing Manis)

Abstract

This research aimed to analyzed longitudinal data after a tree structure regression method being applied to the data, to group some objects with the same response profile. The comparison of mean profile of all groups is also shown, as well as the comparison of each group's data with ungrouped data. The analyzed longitudinal response data are the glucose content of diabetes patients who cured in M. Jamil Hospital, Padang. Explanatory variables which assumed as the ones those have contribution to the response value are age, sex, relative body weight, kind of diabetes mellitus, complication, the recorded length of symptomps appearance and calorie content of patient's diet. The best tree of glucose content has tix terminal nodes, so that based on the glucose content profile, diabetes patient can be classified into six groups. The classification is based on the variables of kind of diabetes mellitus, age, complication, and relative body weight. The comparison applied to confidence band of glucose content mean shows that the groups have different mean glucose content. Futhermore, it is obvious that the grouped and ungrouped data have different mean of glucose content profile. It is also shown that patients who have recognized diabetes without complication and with neuropati perifer have possibility of increase of glucose content during the curing period. For other group, the treatment given gives results as expected.