Considerations when running an experiment like in designing-for-experimental-control:

1. for lengthy experiments, we need to give breaks to the subjects and control the lengths
2. keep people in a lab setting
3. set a time limit?
4. what to record if a student didn’t finish the task?
5. too many dropouts?

Experiments are not just about recording numbers. We need to observe the data qualitatively and make conclusions from that.

Example

Going back to the experiment in designing-for-experimental-control, the formal setup is as follows:

• 40 students
• 2 classes of 20 each
• response: time (mins) -> numeric/continuous/scalar variable
• 1 factor: IDE (Integrated Development Environment)
• 2 levels: Visual Studio (C#), Eclipse (Java)
• this is a between-subject factor.

References

• confounding-variables
• variable-types
• measurement-errors
• measuring-results
• anova-assumptions