Factorial
Sense it. Test it. Improve it.
Design, Run and Learn from your experiments. Nothing to install. Data never leaves your machine.
Your studies (0), kept in this browser
Saved as you work, on this device only. The 30 most recent are kept.
What will you measure?
| Units | Goal | Target | Remove | |
|---|---|---|---|---|
A study measures at most 5 responses.
The first response sizes the design, so the page asks what you know of it.
What you know of it already
What might change it?
| Factor name | Type | Low | High | Units | Remove |
|---|
Settings you cannot run
A mixture takes no limits on the settings on this page: give each component its lower and upper proportions instead.
Type each as an inequality in the factors' own units, with + - * / and ** for a power.
A study can have at most 8 limits.
The settings you can run
Or mark the settings you cannot run
Tick the corners that cannot be run, and a limit that rules them out is proposed.
What do you want to learn?
Designs that can learn this
Enter two factors or more to list them.
The design
Each cell is the of the best fraction in that many runs: going down costs runs and buys resolution.
Show the Python code
The process-improve code for this design. Copy it and it runs as it is.
Run the experiments
No run sheet yet: make it in Design.
The design has changed since this run sheet was made.
Show the true process
In coded units. The analysis estimates these coefficients from the noisy results.
Show the Python code
The design and its results together, so it runs as it is.
What the results say
No results yet: enter them in Run, or practise with simulated ones.
The best settings for your goal
Test them with
Show the Python code
The design, its results, their analysis and the best settings together, so it runs as it is.
Show the Python code
The design, its results and their analysis together, so it runs as it is.
Balance the responses
Each response with a goal scores a , d, from 0 where it is not acceptable to 1 where it is as good as it gets. Its range says where: to maximise, 0 at the low end and below, 1 at the high end and above; to minimise, the other way round; for a target, 1 at the target and 0 at either end. says how much each counts in the , D, and the how fast d rises towards it.
Test them with
Show the Python code
The design, its results and the balance of the responses together, so it runs as it is.