- PII
- S0132-16250000352-3-
- DOI
- Publication type
- Article
- Status
- Published
- Authors
- Volume/ Edition
- Volume 404 / Issue 12
- Pages
- 42-54
- Abstract
- This article addresses the problems of quality assessment in mass surveys’ primary data during the process of «work files» preparation for solving multidimensional analysis problems. Such problems include introduction of quality criteria, some of which are of general nature, applicable to different methods of analysis, and some are specific and dependent on the proposed research practice of analysis. As an example of the latter, typological analysis for the purpose of reconstructing of social types as objects of functional management is being examined. Based on primary survey data of industrial workers as an example, the quality of variables and objects for the purpose of typologization is assessed. The logic of the «battle» with the missing data that presupposes a subjectoriented approach is also proposed here. It is based on the idea of restoration (imputation) not of the original values for the analysisof variables, but of indices that are playing, in particular, the role of classification characteristics. These are the variables that are fed to the input for procedures of separating objects into classes. Several steps are identified in the process of work files accumulation for typological analysis: evaluation of the quality of variables that correspond to the first part of the type-forming characteristics and are used to define and establish classification characteristics (step 1); refinement of the factor structure of these variables (step 2); formation of classification characteristics (step 3); assignment of values to the indexes for missing data cases (step 4); inclusion of variables from the second and third part of the generative attributes in the work file (step 5).
- Keywords
- data quality, typological analysis, methods of data imputation, local data imputation, type-forming attributes, classification characteristics, indices
- Date of publication
- 01.12.2017
- Number of purchasers
- 4
- Views
- 559