Mohammad Asghari Jafarabadi, Seyede Momeneh Mohammadi,
Volume 14, Issue 3 (3-2015)
Abstract
There are situations in medical studies, wherein it is impossible to use the methods based on normal distribution (parametric methods). This paper objects to introduce common nonparametric methods and the inferences based on the methods in medical studies. Principles and method of calculations along with the software codes for common nonparametric methods and inference based on them were presented taking into account the considerations relevant to choose the nonparametric methods and their relative efficiency with examples in medical studies. In the situation where the assumptions are not satisfied, the nonparametric methods should be used without caution to lose the efficiency or even with higher efficiency of these methods. To compare a non-normal or ordinal variable between two groups Mann-Whitney test, to compare a non-normal or ordinal variable among more than two groups Kruskal-Wallis test, to compare a non-normal variable between two related situations or matched groups Wilcoxon test and to compare an ordinal variable between two related situations or matched groups Sign test should be used. In each of these tests the results of research based examples were presented along with the methods of their calculations. To assess the relation or difference in all types of medical studies, these tests are recommended considering the situation and purpose of study.
Fatemeh Dekamini, Mohammad Ehsanifar,
Volume 21, Issue 4 (10-2021)
Abstract
Background: Diabetes is one of the major health problems in Iran and about 4.6 million adults suffer from this disease. Poor diagnosis of this disease has caused half of this number to be unaware of their disease. In recent years, along with the use of computers in data analysis and storage, the volume and complexity of data has increased dramatically.
Methods: In health organizations, data play an essential role in the value of the organization. Therefore, data mining has become one of the most widely used processes in the field of health and disease diagnosis. In this study, the information of 768 laboratory clients in Tehran was kept confidential and the opinions of experts were used to identify the variables affecting the incidence of diabetes.
Results: The findings indicate the study of 5 algorithms on the presented data, which by implementing 5 data mining algorithms J48, Bayes, Beginning, Cohen and simple clustering to classify the data, the efficiency of these algorithms in terms of speed and accuracy in calculations was evaluated.
Conclusion: The data set for classification is the database of a laboratory, which includes 768 samples with 9 characteristics. Finally, J48 algorithm is recommended for data mining of diabetes due to high speed, acceptable accuracy and lack of sensitivity to raw data.
