cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
  • Image to Table JMP extension imports single or batch images as data tables. Useful to extract curves from pictures or convert image data to curve data to use in analyses like Functional Data Explorer. Download and install the extension.

Discussions

Solve problems, and share tips and tricks with other JMP users.
Choose Language Hide Translation Bar

How to graphically present full factorial results

I have completed a full factorial design with 3 variables and 2 modeled responses (in duplicates). That gave me 54 runs in total. Now I have responses, but how do I present my results graphically? If it was RSM, I would have plotted surface plots. But with full factorial, I am at a loss.

4 REPLIES 4
Victor_G
Super User

Re: How to graphically present full factorial results

Hi @RandomCheetah77,

Welcome in the Community !

If you have realized a full factorial design, that may imply you're interested in detecting and estimating main effects and interaction effects. Some idea to plot your results and convey your findings:

  •  Show the model performances graphically to better assess the validity of the modeling: for example through R² and R² adjusted for explainative performances (% of variation captured by the model), and RASE/RMSE or Normalized RMSE for predictive performances. Examples:
    Victor_G_3-1794903924513.pngVictor_G_4-1794903956521.png
    To do this, simply create a table from Summary of Fit panel, right-click on it and click on "Make Combined Data Table" if you have two responses, this will create a table with the model diagnostics for the two responses (provided the panel are available and open in the JMP window).

  • If you have built several models (with several estimation methods for example), you may be interested into looking at how each model terms are included or not between the models. That may help you determine how confident you can be in the statistical significance of some terms, based on their detection by several models:
    Victor_G_2-1794903823443.png

    To do this, same situation as before, but you can create a table from the Parameter Estimates panel.

  •  You can also display the effect size plot, directly from the JMP model window with the Scaled Estimates panel, or re-create graphically the plot using same techniques as before, right-clicking on the table of expanded estimates (to get all the terms) to create a new table on which you can create graphs:

    Victor_G_9-1794904221497.png

    You can put the emphasis on different parts, the size of the estimates (through a Pareto plot) or the term type (main effects and interactions):

    Victor_G_11-1794905037695.pngVictor_G_10-1794904534542.png
  • You can also look at the Interaction Plots to see how the interactions between factors influence the response(s):
    Victor_G_13-1794905165615.png

Check also the different residual plots available in the Row Diagnostics panel before looking at other informations, to make sure your model is appropriate (and more particularly the absence of pattern due to a misspecified distribution or the absence of some missing terms in the model).

Hope this answer will help you,

Victor GUILLER

"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)
MathStatChem
Level VII

Re: How to graphically present full factorial results

Fit a linear regression model (main effects + 2 factor interaction), and then use the Prediction Profiler.  For example, using the sample data file Reactor 32 Runs.jmp:

MathStatChem_0-1794924297611.png

Also, if you want to visualize the actual data, Main Effects Plots:

MathStatChem_1-1794924397949.png

and variability charts

MathStatChem_2-1794924475004.png

are useful visualizations.

MathStatChem
Level VII

Re: How to graphically present full factorial results

and, for factorial designs, the Cube Plot is an "oldie but a goodie" type of plot for seeing the model predictions.

MathStatChem_3-1794924640894.png

 

if you fit the full factorial model, you can also get a more complete Cube Plot:

MathStatChem_4-1794924728283.png

 

 

statman
Super User

Re: How to graphically present full factorial results

I'm a bit confused, you haven't provided enough context.  You say you have a factorial with 3 factors, how many levels? Are the factors continuous or categorical? Are the responses continuous, nominal or ordinal? I'm still not sure how you got 54 treatments? I'm guessing the 3 factors were tested at 3 levels? You mention "duplicate". I don't know if those are replicates or repeats (yes it matters). You suggest you have 2 response variables. I would want to create a simple scatter plot and test for correlation and outliers first. Then, if appropriate, normal plots, Pareto plots, main effects (prediction profiler) and interaction plots. Simplify the model and there are a host of residual plots to look at.

"All models are wrong, some are useful" G.E.P. Box

Recommended Articles