Import, filter, and pre-process data, and detect patterns

Theme can import, filter and pre-process data, and detect patterns by objective statistical criteria. The resulting pattern sets can be displayed, searched, and analyzed in a number of ways. Pattern statistics can be exported to statistical software packages for further analysis.

Import data and detect patterns
Theme works well on most raw behavioral data. It has built-in pre-processing options for filtering and reorganizing the data so data can be imported easily. When you import data, the pattern detection engine can be tuned to the desired level of detail, sensitivity, types of pattern, and calculation time. Pattern detection works best on continuously scored events, such as the data produced by The Observer XT.

Search and filter the detected patterns
The detected patterns can be searched and filtered in a number of ways, allowing you to focus on important details. There are simple criteria like behaviors, pattern length, pattern frequency, and number of actors involved in patterns. There are also advanced criteria based on the relationships between patterns and the underlying event data. Criteria can be freely combined and the results can be stored for later use.

Presentation of patterns
The overall structure of the behavior is presented in a matrix of linked event types. An overview of the position of event types in patterns gives an impression of their relative importance. At the lowest level of detail, individual patterns can be displayed on a timeline and the associated video fragments can be played to see the pattern unfold. You can use this feature for presentations and conferences.
Theme gives all the details for each pattern: exact times for each occurrence, significance, and various other statistics. For behaviors with a spatial element like choosing items in a store, Theme can show a path graph.

Perform statistical tests and export data
Theme provides a full range of descriptive statistics per input file, such as number of patterns, length, and diversity. These statistics are measures of overall structure which can be more sensitive to the effect of experimental treatment than frequency and duration based statistics. The statistics can be exported as a table to statistical packages, with their independent variables.