STRUCTURE Data Files > Structure Behavioral Events

Doing research basically means exploring something new, so there is no standard system on how to structure your data. Maybe within your field of application there is, but not in general.

Your data structure, has impact on the statistics and the original data can be re-arranged, in order to answer specific questions.

You'll probably discover that you need multiple, differently structured, versions of the same data to answer all questions during analysis.

INTERACT offers many routines for restructuring AFTER you have finished collecting your data!

The most important structural element is the DataSet:

oDataSets, group behavioral Events from the same session, task or person.

oIn statistical calculations, export and graphical representation, the results are split per DataSet.

oOther routines like the co-occurrence filter or sequence analysis are also performed per DataSet.

oTypically, each observational session is stored in a single DataSet.

oMultiple observational passes through the video can be stored in the same DataSet.

Additional DataSets can be created whenever wherever you need them.

Restructuring

During analysis you might discover that some results are hard to get in your original structure. That is why INTERACT offers a range of restructuring commands to easily get the results you need.

Some initial structures are chosen only to make coding life as easy as possible and restructuring is part of the plan to start with.

You probably need multiple versions of the same data to answer all your research questions.

Restructuring Reasons

A few examples:

oMerge DataSets - When coding multiple Subjects in separate DataSets, statistics per Subject are available right away.
Co-occurrences of behavior can be visualized and counted with the State-Space-Grid, but creating new Events for those overlaps with the Co-occurrence Filter or finding Contingencies requires all data to be in a single DataSet.

oSplit DataSet per Code - When coding multiple subjects in a single DataSet (using lexical chains and a subject code per line) statistics per subject are not available straight away.
To compare a single behavior per subject for all subjects at once, use the State-Space-Grid.
To get the statistics of all Codes per subject in a structured way, split your data into a separate Set per subject with a single command.

oInsert DataSet - During analysis, you might discover that splitting your session into different Sets per Phase or question or task would give you better visualisation options or easier to read statistics. If there is not Code for which you can do this automatically, you can insert a Set between line manually. Both methods are described in Split Events into Separate DataSets.

General Structuring Advice

Some common structure advises are:

oLog your observations in a separate file per session, mainly for data security reasons.

oUse a separate DataSet for each set of videos.

oLink your video file(s) to the DataSet that holds the corresponding Events.
See Link Video Files for Analysis

oUse the DataGroup element to group a collection of DataSets.
For instance, a DataGroup can be used per subject, and each DataSet can represent a different session. The other way around works just as well: e.g. grouping different subjects per Task.

oEnter a description into the DataSet and Group description line for easier identification.
See Identify DataSets and Groups.

oMerge multiple documents before analysis or other adaptive routines.
See Merge Documents.

oUse Independent Variables for easier data selection during analysis, especially if you have a large number of DataSets.
See Assign Variables.

oFor large coder groups, define a document template that holds a predefined DataSet structure, available independent variables, Class columns and reference to the CodeArchive. This minimizes inconsistencies.
See Work with Templates.

Note: Some analysis related corrections ask for moving and/or combining Codes and inserting new Classes, to move or copy those Codes to. Other goals can only be achieved by splitting your Events into multiple DataSets.