Statistical Discourse Analysis: Testing Educational Hypotheses with Large Datasets of Electronic Discourse

Abstract

Educators are increasingly using electronic discourse for student learning and problem solving, partially due to its time and space flexibility and greater opportunities for information processing and higher order thinking. When researchers try to statistically analyze the relationships among electronic discourse messages however, they often face difficulties regarding the data (missing data, many codes, non-linear trees of messages), dependent variables (topic differences, time differences, discrete, infrequent, multiple dependent variables) and explanatory variables (sequences of messages, cross-level moderation, indirect effects, false positives). Statistical discourse analysis (SDA) addresses all of these difficulties as shown in analyses of social cues in 894 messages posted by 183 students during 60 online asynchronous discussions. The results showed that disagreements increased negative social cues, supporting the hypothesis that these participants did not save face during disagreements, but attacked face. Using these types of analyses and results, researchers can inform designs and uses of electronic discourse.