A MODEL FOR COMPUTER-ASSISTED ANALYSIS OF OPEN-ENDED SURVEY AND QUESTIONNAIRE DATA.

摘要

Problem Statement. This study evaluated the appropriateness of text processing methods for analyzing uncoded, open-ended survey and questionnaire data using a computer. The methods evaluated were organized into a three-dimensional model, focusing on: number of variables, including univariate, bivariate and multivariate methods; classification process, including clustering and partitioning methods; and analytic method, including descriptive, associative and probabilistic methods.;Findings and Conclusions. Several of the model's cells were found to be appropriate for analyzing three open-ended survey data sets. Among them were univariate and bivariate clustering and partitioning methods. Multivariate methods were only appropriate for the data set with the longest responses. Results from the computer-assisted analysis were compared with results from a manual analysis for one of the three data sets. Themes identified with both methods were comparable. Since length of the original responses determined which analytic techniques yielded the most interpretable results, the model was reduced to two dimensions: length and number of variables used.;Methodology. A computerized text processing system was developed permitting the implementation of the analytic techniques within each of the model's cells. The system, which processed responses in their original form, included: a stop word list of high frequency words eliminated from responses; four synonym dictionaries used to equate equivalent words; procedures for normalizing the lemmatizing words; and procedures for applying the analytic techniques.