topic Re: Questions about DOE in Discussions

Questions about DOE

AHM — Thu, 23 Nov 2023 10:03:41 GMT

Hi everone

I would like to conduct a field trial. I would like to check the response of a new fertilizer (at different levels) with and without exisiting fertilizers on yield of wheat.

i need help with DOE. so basically i have eight treatments T0= negative control (no fertlizers), T1= positive control (NPK) business as usual fertlizers, T2= NPK+low level of new fertilizer, T3=NPK+medium level of new fertlizer, T4=NPK+high level of new fertilizer, T5=low level of new fertlizer, T6=medium level of new fertilizer, T7=high level of new fertilzier.

i need to check the effect of treatments on yield. i have planned 3 replicates or blocks (Randomised complete block design).

how can i use DOE to design this trial in JMP.

I shall welcome all the suggestions.

Thanks

Re: Questions about DOE

Victor_G — Thu, 23 Nov 2023 10:40:06 GMT

Hi @AHM,

Welcome in the Community !

Regarding the treatments you would like to have, why not considering a DoE with two factors :

X1 : Quantity of NPK (with levels at 0 and at "normal/standard" quantity), numeric continuous factor
X2 : Quantity of new fertilizer (with levels at 0, middle level and high level), numeric continuous factor.

Using a standard factorial approach, and specifying a model with the two main effects, a quadratic effect for X2 to have experimental runs with 3 possible levels ( if you want to have 4 levels for factor X2 : absence, low, middle and high levels you can specify a cubic order effect term in addition in the model) and the interaction between these two factors, you should have all the treatments you are expecting :

You can create this Custom Design with this script (you can modify the quantity levels used, I just used them for illustration):

DOE(
	Custom Design,
	{Add Response( Maximize, "Y", ., ., . ),
	Add Factor( Continuous, 0, 5, "NPK Quantity", 0 ),
	Add Factor( Continuous, 0, 10, "New fertilizer quantity", 0 ),
	Set Random Seed( 14871954 ), Number of Starts( 312796 ), Add Term( {1, 0} ),
	Add Term( {1, 1} ), Add Term( {2, 1} ), Add Term( {2, 2} ),
	Add Alias Term( {1, 1}, {2, 1} ), Set Sample Size( 8 ), Simulate Responses( 0 ),
	Save X Matrix( 0 ), Make Design}
)

And datatable of the DoE (without replicates) is attached.

If you want to replicate the whole design 2 more times, you can use the "Augment Design" platform (in menu "DoE", "Augment Design"), click on "Replicate" option and specify 3 as the number of replicates :

You'll end up with a completely randomized design with 3 replicates, for a total of 24 runs (datatable attached).

If you have constraints about number of runs done per day for example, you can use blocks in the initial design generation, by adding a factor (type blocking) and specify the number of runs per block :

You can then specify the same model as before (with added block effect), the number of runs in total (24), and you'll end up with a design in 3 blocks (see datatable attached).

I hope these design options may help you in the creation of your DoE,

Re: Questions about DOE

AHM — Thu, 23 Nov 2023 13:20:28 GMT

Thanks a lot @Victor_G

As i am new to jmp and statistical world here please do apologize my stupid questions

1. I can rephrase my question again

Factor 1 (NPK) (2 levels)= 0, NPK

Factor 2 (GHQ) (4 levels)= 0, 5, 10, 50

Replications = 3

Treatments

T0= control (no fertlizers)

T1= control (NPK),

T2= NPK+5GHQ

T3=NPK+10GHQ

T4=NPK+50GHQ

T5=5 GHQ

T6=10 GHQ

T7=50 GHQ

I wish to have RCBD which will give me treatment * replication = 8*3 = 24 plots

Can you please help a DOE for this setup? Please suggest a script to run

for attached files you have sent, how can i go to their scripts if i want to edit them?

Looking forward for your kind response

Re: Questions about DOE

Victor_G — Thu, 23 Nov 2023 13:54:47 GMT

With your informations, here is how the setting of the factors and model can be done in order to get the 8 treatments you need :

To obtain this 8-runs design, here is the script :

DOE(
	Custom Design,
	{Add Response( Maximize, "Y", ., ., . ),
	Add Factor( Categorical, {"0", "NPK"}, "NPK", 0 ),
	Add Factor( Discrete Numeric, {0, 5, 10, 50}, "GHQ", 0 ),
	Set Random Seed( 1540555554 ), Number of Starts( 329553 ), Add Term( {1, 0} ),
	Add Term( {1, 1} ), Add Term( {2, 1} ), Add Term( {2, 2} ), Add Term( {2, 3} ),
	Add Term( {1, 1}, {2, 1} ), Set Sample Size( 8 ), Simulate Responses( 0 ),
	Save X Matrix( 0 ), Make Design}
)

Then, to have 3 replicates, you just need to use this design and go to "Augment Design", option "Replicate", specify that you want each runs done 3 times, and you'll obtain your 24 runs as needed (datatable attached). The script used to augment the design is :

dt = Current Data Table();
dt2 = dt << run script( "Original Data Table" );
Current Data Table( dt2 );
DOE(
	Augment Design,
	X( :NPK, :GHQ ),
	Y( :Y ),
	{Augment Method( Replicate, 3 ), Save X Matrix( 0 ), Simulate Responses( 0 ),
	Make Table}
);

Note that you can find these scripts and modify them in every DOE datatable, by right-clicking on the green triangle next to "DOE Dialog" and then clicking on "Edit". This will open a windows with the script used to generate the design.

Hope this answer will help you,