Seven Steps of an Experiment
- Recognition & statement of problem.
- All members should agree on.
- Choice of factors, levels, and ranges.
- Teamwork!
- Experts tend to design experiments that are too narrow in order to prove their preconceived answer to the problem.
- Don’t let that happen! Bring in other points of view.
- Choosing the factors
- Levels: fairly low numbers.
- Ranges: be aggressive enough to see the actual impact instead of noises.
- Teamwork!
- Selection of the response variables.
- What’s the metric?
- How to get those metrics? Well-calibrated instruments?
- Choice of design.
- Conducting the experiment.
- Do it yourself!
- Be there! Make sure the experiments are done correctly.
- Statistical analysis.
- Easy as long as steps 1 through 5 are done correctly.
- Drawing conclusions, recommendations.
- Data visualization matters.
Takeaways
- Get statistical thinking involved early.
- Non-statistical knowledge is crucial to success.
- Pre-experimental planning (1-3) is vital.
If you have ten weeks to solve a problem, you should spend eight weeks planning the experiment, one week executing the experiment, and one week analyzing the data.
- Think and experiment sequentially (KISS principle).