Hi Sanghee,
The size of the dots is 48*48px as of PennController 2.0. The calculation starts immediately when the middle dot shows up, and the score is proportional to the average distance from the center of the screen over the (X,Y) coordinates of the estimated gazes (precision = 100 - (distance / halfWindowHeight * 100)). As far as I remember, and as far as I can tell now, there is no default threshold value: if none is provided, then all calibrations will be successful. Reference
There is no set sampling rate: if memory serves, the WebGazer library will try to run at each update of the visual frame (look up requestAnimationFrame) which will happen more or less frequently depending on the performance of the participant’s browser at the time of the experiment.
Regarding the Latin Square design, this is something you need to code yourself in your CSV. PennController picks one value from the ones listed in the “Group” or “List” column and subsets the table to all and only the rows containing that one value in that column, and generate the trials from those rows. If you want to cycle through conditions across items in Latin Square fashion, you need to design your table accordingly, as illustrated in the advanced tutorial
Jeremy