تحلیل مشاهدات گمشده در مطالعه اثر دوزهای مختلف مکمل ویتامین D بر مقاومت به انسولین در دوران بارداری

Introduction: The aim  of  this  study  was to impute missing data  and  to compare the effect  of  different doses of  vitamin D supplementation on  insulin resistance during  pregnancy. Methods: A clinical trial  study   was done on 104  women  with diabetes and gestational age less than 12 weeks between 1391 and 1393. These   subjects  were   randomly  divided  into  three  groups; pregnant women who received  daily 200 IU vitamin D (group A), women where receiving monthly 50,000 IU vitamin D (Group B) and Group C are women who received 50,000 IU vitamins D every two weeks. In order to investigate the effect of missing data, the data were studied in two ways, with and without considering missing data. To analyze data in the presence of missing observations, the mechanism of MCAR is considered as the missing mechanism. Then, in order to impute the missing data, four methods including mean imputation, random overall hot-deck imputation, within-class random hot-deck imputation and nearest neighbor imputation was used. Results: In this study, in random overall hot-deck imputation, the difference between blood sugar and insulin resistance variables are not normal, so median and their interquartile range were reported in the table. Furthermore, kruskal-wallis test was used to compare 3 groups variables. The difference   insulin resistance variable was not normal in the nearest neighbor imputation method, so the median and interquartile range was reported in the table. In addition, the kruskal-wallis test was used to compare 3 groups of data. The delta index was calculated for all imputation methods. Conclusion: In this study, delta index was calculated to evaluate and to compare imputation methods. The random overall hot-deck imputation was described as the best imputation method.