Problems performing a DOE

[#1]

Hello,

I am trying to solve different DOE exercises. However, JMP will not really let me perform this DOE. I have solved 4.1 - 4.4, but have difficulties from 4.5-4.7.

Any suggestions or hints?

4.1:

We have four factors with 8 runs, and we then have a fractional factorial with a resolution of 4.

4.2: 

We have 4 factors: LG, TG, offset, and tube.

4.3:

• Fractional factorial experiments are good to find main effects and the interactions. The resolution of the experiment is 4 which means that main effects are confounding with 3 factor interactions and other 2 factor interactions are confounding with 2 factor interactions.
• You can consider the resolution of a design to be an indication of how clearly the effects can be separated. In a fractional experiment you sacrifice not knowing where the impact comes from due to confounding.

4.4: 

• The experiment does not have centerpoints, therefore we cannot check for linearity.

[#2]

Welcome to the community.  Is this all of the information you received?  How much of a change in the OD matters? Why didn't you attach a JMP file instead of an Excel file?

At first glance, it looks like you have 3 factors in an experiment and the tube thickness appears to be a covariate but there is not sufficient explanation to determine this (was this factor specifically manipulated or was it simply measured for each treatment).

If we assume the tube was manipulated, then the res is IV as you note.

To answer question 5-7, run: Analyze>fit model.  You can start with a saturated model and then reduce the model removing insignificant terms.  Re-run the model and analyze the residuals (many plots to help with this). Save the prediction formula and solve for 2.5 OD.

[#3]

Thank you very much. Ah I didn´t think about attaching the JMP file instead - sorry about that.

Yes that is all data available for the exercise. Ahhhh, I think you are right regarding the factors.

Now I did a full factorial with 3 factors and deleted the 3 factor interaction to see the PValues. I know the PValues should be below 0,05, but that is not really the case here, and it doesnt really help to delete more. Would you continue reducing the model, or just go with this result?

Thank you so so much, your answer is highly appreciated.

[#4]

I'd say the bottom 3 effects in your effect summary list should be removed from the model.

I'm guessing they want you to keep the tube thickness in the model and perhaps an interaction of tube thickness and another factor (based on the question in the list).  What I would do is saturate the model (4 factors and 3 2nd order effects.  Use normal plots and Pareto plots as well as SS, to reduce the model.

[#5]

I have tried again and again, and now I actually think I have solved everything except 4.5. I have attached the JMP file.
I dont know if it is me who is stupid, but I am really not sure if the linear model should be an equation or just a graphical linear model, if you know what I mean?