Restore of the tables with partially missing data

Abstract

The problem of data recovery found in a wide variety of statistical analysis applications. For example, it related to the survey in various sectors of the economy. When filling out a questionnaire specialist specific subject area sometimes there is a situation where he cannot answer all the questions for technical reasons (failure of the communication channel) or for personal reasons (he may not prefer one criterion to another). Survey data in many cases are formed in the form of rectangular binary tables, where columns (rows) are the studied variables that correspond to binary numbers (1, 0) or signs ( + , -). This article is devoted to the analysis and recovery of binary data in cases when the table of values of variables for the above reasons is absent. The solution of this problem proposed to carried out by the method of homogeneous groups. The essence of this method is to find in the General population of binary data of homogeneous groups of subjects of the survey and determine the affiliation of the subject of the survey, in the response of which for one of the above reasons there was a transition to one of the selected homogeneous groups. This method works well for large amounts of binary data with relatively few spaces. Otherwise, stricter methods, such as simulation methods using maximum likelihood functions or Bayesian strategies, are used. Ref. 8, tabl. 2

Keywords

• method of questionnaire
• questionnaire table of binary data
• homogeneous data groups
• the method of homogeneous groups
• restoration of missing data.