Data exploration was done to get a feel for the overall dataset, test for normality and look at outliers or predominant trends across variables. My first step was to create simple box plots to identify trends across age and food treatments, leaving sex out of the quation for the time being as this would have gotten too messy at this point. The first trend was that there seemed to be changes in mean behaior with increasing age. Food seemed to have an impact too but was in its effect more subtle and specific and wasn't quantifiable just by pure data exploration. However these trends supported me in my assumption that quantifiable food, stage and possible sex affects would be apparent in my model results. In the end I was also left with the choice deciding between using time in inner area (boldness) or time in outer area (shyness) and decided to use shyness as a proxy for the actual analysis but both expressed behaviors are interchangeable.
- Taking a look at overall repeatability -
The second step was related to my second research question and using simple point and line plots allowed me to investigate repeatability trends across stages, food treatmens and trials to make first assumptions if actual behavior is measured or just random variation. Despite individual outliers, the majority of fish behavior seemed constant across trials supporting the overall experimanetal setup assumption of quantifying distinct behaviors with the open field test.
- Data normality and distribution -
As seen normality was not given for most of the behavioral response variables (Time in outer area, time spent moving, actual velocity). Thus data transformation was performed (arcsin and log).