Graphed Interactions in Sequence-space from Time-series - GIST

The Real-time Instructor Observing Tool (RIOT) is an application that allows quantitative time-series data from classroom interactions to be recorded and analyzed. The application displays a wealth of information from a single classroom observation in a time-dependent color-coded plot, which visually represents the prominent features of a classroom in an immediately accessible way, greatly simplifying initial analysis. While the representation allows for quick conclusions, it does not preclude a more in-depth analysis. To accomplish this we created Graphed Interactions in Sequence-space from Time-series (GIST) to produce a deeper analysis with the goals of: (1) finding known interaction sequences (IS) and analyzing the sequences preceding and following them; (2) to identify unknown IS; (3) to create a means of visualization of the time-series data (TSD). Categorical TSD present a particular challenge for analysis. Visual representations of data can be difficult to interpret except when limited TSD categories are displayed. We have created a way to visualize TSD through a combination of unique analysis and graphing that will benefit researchers in a diverse range of fields. Inspired by “shotgun” sequencing of DNA the GIST algorithm breaks TSD into a spectrum of IS with a user defined range of lengths. The spectrum is searchable for a specific interaction sequence (SIS). Sequence counts, and durations of interactions are used to produce a modified node graph with weights assigned to nodes and edges such that the normalized occurrence and duration of each IS can be easily seen. With antecedent and response sequences attached to the centrally placed search sequence, it is easy to see the dominant patterns of interaction both leading to the SIS and resulting from it. Using GIST to analyze RIOT data collected from different instructors at different times, we searched for and found the alternating three interaction SIS of Clarifying Instructions to the Whole Class (WC), a Small Group (SG), then the Whole Class again. The SIS had an even distribution of response IS. However, the antecedent IS were not evenly distributed; the WC to SG IS for Explaining, Listening to Question, and Ideas each having double the occurrence of other antecedent IS. A pattern of instructors utilizing WC then SG interactions of these types before the SIS is clearly present. Further data analysis using GIST will identify the dominant IS for individuals and groups. Relationships between antecedent IS, SIS, and response IS may provide insight into what drives particular IS to occur.