7.4. Data analysis

Experiments with the temperature arena produce a large amount of data in the form of temperature logs and video recordings. A suite of software programs is available to reduce the analysis effort to a minimum:

  • Freely available video editing software is used to extract individual frames from the video recordings at a deliberate rate.
  • A MatLab® script (available at http://zool33.uni-graz.at/artlife/sites/default/files/track_single_agent.zip) iterates over these frames, identifies and extracts the positions of single bees or groups of bees and stores them in a Microsoft Excel® file.
  • The number of individuals in a group can be determined with sufficient accuracy from the area covered by the group.
  • A Microsoft Excel® script (available on request from http://zool33.uni-graz.at/artlife/contact) gathers the sequence of positions from all experiments, reads the temperature logs and links the positions to the local temperature.
  • The script establishes the trajectories of single bees (e.g. Fig. 10) and analyses a number of behavioural parameters (e.g. distribution of moving speeds, turning angles, moving and resting behaviour or the time spent in distinct arena zones like optimum, sub-optimum or wall zone) in dependence of local temperature and gradient steepness.

This analysis can serve as a bio-assay to assess the health status of a colony by comparing the clustering performance of a sample group from the infested hive to the performance of a comparable group from a healthy colony (for data analysis see the BEEBOOK paper on statistical methods (Pirk et al., 2013)). The clustering performance is determined by the time it takes until a given percentage of the individuals are clustered in the optimum, the ratio of individuals in the optimum after a given time and the distribution of individuals between the global and a local optimum. Groups from colonies with a compromised health status are expected to have a decreased clustering performance, which can be detected using established statistical methods.