JMP CUSTOM DESIGN

[#1]

Good day, Mark Bailey @Mark_Bailey ,
Please I have an issue setting up a custom design for my experiment.
According to the design experiment, it is supposed to be a Split plot design (having two drought treatments, with each having 3 XYZ-crop genotype replicated 4 times giving it 24 runs in all I believe). However, I am finding it difficult to set it up in custom design or should I say I'm not confident enough with what I did because I am new to JMP.
This is what I did:
I set my response (which is Chlorophyll), Factors as Treatment (Hard): two levels(WL-water limited and WW-Well watered) and Genotype as Easy (3 levels: x, y, z), Set whole plot as 2 and runs as 24 but I'm in doubt of how to set in the replicates. Please I need your assistance on the right path. Thank you

[#2]

First of all, you needn't address a discussion to me. I will answer if I can. But you might inhibit other community members from replying with potentially a better idea than mine!

Why is the treatment hard to change? How will you run the experiment?

Also, if you have two levels of the hard-to-change factor and you request only two whole plots, you will lose some estimation because JMP will have to steal degrees of freedom from the model itself. Can you use 4 whole plots or more?

[#3]

Oh noted, I will endeavor to make it open to all next time.

I am really not sure of how to set the nature of my experiment up in JMP for proper analysis. Like earlier posts I have 3 genotype of a crop having 4 replicates each, subject to two different drought treatments. (Main factors: Drought treatment and Genotype). I have attached a view of my field map. Green represents the Well water and Red represents WL.

Thank you 

 

[#4 · Accepted solution]

Well, I wish you were using more than two whole plots, but I understand that you have a practical constraint.

Here is a custom design definition that I think captures the essential characteristics as you explained them.

This design is the result.

The warning in red is by the custom design. Here is a script that will re-produce the custom design platform for you to study and possible use. You can copy it here and paste it into a new script window.

DOE(
	Custom Design,
	{Add Response( Maximize, "Chlorophyll", ., ., . ),
	Add Factor( Categorical, {"Water Limited", "Well Watered"}, "Drought", 1 ),
	Add Factor( Categorical, {"x", "y", "z"}, "Genotype", 0 ),
	Set Random Seed( 809122516 ), Number of Starts( 13501 ), Add Term( {1, 0} ),
	Add Term( {1, 1} ), Add Term( {2, 1} ), Add Term( {1, 1}, {2, 1} ),
	Set N Whole Plots( 2 ), Set Sample Size( 24 ), Simulate Responses( 0 ),
	Save X Matrix( 0 )}
);

[#5]

Thank you so much.

Would the model still remain valid if I stack the results of the chlorophyll response gotten over a span of days?

Like following the model design; I have Chlorophyll for days 1, 2, 3 ..... n, is it possible to stack them and have another variable Date still building it from the model source?

Thank you...

[#6]

There is more than one way to incorporate the time element of your study. I would not include Day as a factor. Why? Because custom design is model-driven. Let's say that your study was 10 days long and you were to record the response each day. That situation would require using a Discrete Numeric factor in which you would enumerate the 10 days. That would force JMP to enter Day in the model with all terms up to the power 11!

Instead, I suggest that you record the daily measurements in additional columns to the right of the Chlorophyll data column. You can add as many columns as needed for the number of daily measurements you make. This way gives you a lot of flexibility depending on your purpose. Why are you measuring on multiple days? Do you intend to model the rate of growth? Do you intend to use the maximum response? Do you intend to use the final measurement at the conclusion of your study? It depends...

[#7]

I intend to model the chlorophyll response across the growing season which I have divided into 9 measuring dates. 

[#8 · Accepted solution]

I would use the approach of adding nine data columns to record the data. You can stack these columns for the analysis and add the date as a continuous numeric factor. You can cross date with the other factors to determine if the rate depends on the conditions.

[#9]

Thank you. This works

[#10]

Good day, would using a repeated measures analysis work for this seeing that I have two factors in the experiment
with response (Chlorophyll) measured over time? Like running a MANOVA for it.