Full factorial with complex continuous factors

[#1]

I'm a n00b who is trying to run a full factorial. i'm just learning JMP and T&E so i don't really know what i'm doing yet.

I have 4 continuous factors. Let's say, my factors are birds observed in X park (where X represents a different local park). i don't know how I should do the levels of the factors. if I want that number to be randomized between 0 and 50, how should i run the test?

if i run the FF with 4 continuous factors and 2 levels (0 and 50), JMP only runs combos of 0 and 50. but technically, any number between 0 and 50 could appear. do i create 4 continuous factors with 50 levels so that each number between 0 and 50 appears? is keeping the 2 levels the best thing? should I do a custom DOE and make the levels binned by fixed numbers?

i also ran a FF with 4 categorical factors (4 levels: none, low, med, high). 

[#2]

Hello @jmpbeginner843,

Welcome in the Community!

I'm sorry but you're giving very little information to help you. Here are some preliminary questions to better understand your topic :

• What is your objective with this DoE ? Prediction, optimization, understand relationships between variables ?
• What are your factors ? You're mentioning 4 continuous factors, but you only mention one linked to the number of parks. Are the factors independent? Or are some levels of different factors linked/constrained ?
• What is/are your response(s) ? Do you know, or can you assess, the experimentation and measurement error ?
• Why did you choose a full factorial design in the first place ? I think there might be better alternatives to reduce the number of runs (with fractional factorial designs, optimal designs, ...) or to explore more uniformly the design space (with space filling designs). Full factorial designs and classical factorial designs use only two levels for continuous factors. The number of levels for continuous factors in factorial designs is linked to the assumed model you have : with 2 levels for your continuous factors, you expect a linear model with first order terms (main effects), and possible interactions. If you have 3 levels for these factors, then your assumed model can detect quadratic effects in addition to previous effects. You can still expect some curvature in your response surface if you have detected strong interactions, but if you're interested in sampling uniformly in the design space, it might be more appropriate to use space filling designs.

Have you checked the different DoE ressources available and the Easy DoE platform to help you get started ?

Hope this first discussion starter may help you,