Calculate constrast and effect for each factor to define which one is significant factor

[#1]

Good evening everyone,

I have 5 factors (A,B,C,D,E) in this excel. However, the only haft-fraction of a 2^5 experiment is used to study. The excel file is provided with test run and observation. I used taylor (1/ -1) to fill the design matrix for main effect with interaction effects.

Based on this excel value, I need to find the contrast and effect for factor A,B,C,D, and E.

I did this in excel:

    Contrast function = sumproduct (A2:A17,F2:F17)

    Effect = constrast (A)/0.5 number of trials.

If you would help me how to calculate the effects and abstract for all 5 main effects based on 16 values provided in excel. I would really appreciated. 

Excel file showed how I use excel function for effect and contrast. This is my senior projects and I have to use jmp.

[#2]

Welcome to the community.  I'm struggling a bit to provide specific advice since this is related to your educational development.  Assuming you know how to open the Excel file in JMP, I suggest you start here:

https://www.jmp.com/support/help/en/16.0/#page/jmp/overview-of-the-fit-model-platform.shtml#

[#3]

Hello,

I have learned about JMP, but in this case, for the effect , I only have 16 effects instead of full factorial 32. Suppose, 5 factors will have 32 effects. I wonder if is there anyway I can still run the DOE full factorial design without any error in Pvalue to get the effect summary. Thank you 

[#4]

Again, I'm not sure from an educational perspective if you are supposed to know this or learn this?  You will not be using a full factorial. 5 factors in 16 treatments is a Res V ½ fraction (2^5-1).  The model should contain all 1st and 2nd order effects (you can create this in JMP by running the Fit Model platform, highlighting the 5 factors and selecting Macros>Factorial to Degree and use the default- 2.  This will saturate the model (account for all of the degrees of freedom). You will get ANOVA and Effect Tests, but there will be no estimate of error and therefore no p-values ( I would use Normal and Pareto plots along with scientific/engineering knowledge to reduce the model).  Once the model is reduced, you can re-run it and the terms that were removed from the model will be pooled into the error terms.  Then you will get the glorious p-values (which, of course, is biased).