The Processing And Applying Of Data Missing Of Radon Measurement By Activated Carbon

Since radon and its daughters were successfully collected in the 1920s, Radon measurement by activated carbon considered as one of the most ideal static, accumulative method, has been widely popularized and applied in many fields. The date of radon measurement is as the main basis of deducing detection results, its quality and correct explanation is particularly important.With the needs of application and the develops of research, A giant amount of research and experiments have been done about the data processing technology of radon measurement by activated carbon in correction of influencing factors, the analysis and denoise of spectral lines, the slippery processing of interference factors, trend analysis and mapping interpretation and so on. Missing data, however Stationing acquisition in the wild and data extraction indoor, is a widespread problem and are completely ineviTable., there is still less research in this question. Missing data not only make further data analysis work becomes more difficult and complicated, and will make the treatment results appear bias in different degree, thereby affecting the precision. The traditional processing method has certain limitation in dealing with a lot of areal measuring radon data. Therefore, we need to choose other appropriate filling method for the missing data and research them.Compared with the research on missing data in our country, foreign research started earlier. At present, several methods were commonly used in our country: Expectation Maximization, Regression and the most popular Multiple Imputation and so on, and various methods have their respective characteristics. In order to study the feasibility, the treatment effect and applicable range associated with the three methods—EM, regression and MI method—in different missing case, based on the study of the basic features of measuring radon data, this article created 15 different mathematic model divided into two missing pattern categories. Through the corresponding processing results coming from various model simulation processing based on the same original model, we find that:In arbitrary missing pattern, processing results will appear larger bias until the rate of missing is as high as 55%, all imputation efficiency is really good when the rate of missing below 35%. And for monotone missing pattern, processing results will appear larger bias and have a major influence to data, when the rate of missing is just 25%, so we should to avoid or have to field completion.This paper mainly used statistical software SPSS, EXCEL, Surfer and so on, by their strong function of data processing and drawing, to relatively research different results, and proved its practical application in the example. The results of the study show that this paper used these methods is feasible in processing a lot of areal measuring radon data, and the results were satisfactory, they all have certain advantage over with traditional methods. At the same time, these methods be used in processing missing data of transient electromagnetic method is also feasible. Especially EM method fully embodies its stability and superiority in the study, followed by regression method, but the present more promising MI did not show more advantages. Thus we can know, for different characteristic of data, the choice of method is different, and the three methods are all feasible in processing missing data of measuring radon and similar areal data such as transient electromagnetic data.