I'm currently working on a Design of Experiments (DOE) for a process that involves temperature shifts at three different times. My challenge is incorporating a 'No Shift' scenario as one of the factors. My initial idea was to: * Treat the three time points for temperature shifts (1 hour, 2 hours, 3 hours) as a continuous factor. * Introduce a 'No Shift' scenario as a categorical factor (yes/no). However, I'm struggling with how to implement this without creating illogical combinations of 'No Shift' with a time point. My questions are: * Is there a way to set up a mixed-level factor that combines continuous and categorical levels without creating invalid combinations? * Are there alternative approaches that would be recommended here? Thank you very much!

Hi everyone, Thank you for your responses and questions. Let me clarify my experimental setup and objectives to provide a better context: * Experimental Setup: In my bioreactor process, I'm examining the effect of a temperature shift from 37°C to 35°C on antibody production using mammalian cells. The shift can occur at different times: 1 hour, 2 hours, or 3 hours into the process. However, I also want to include a scenario where this temperature shift does not happen. * Objective: The primary purpose is to maximize the yield of the antibody. I aim to understand how the timing of the temperature shift impacts the yield and how the yield compares when there is no temperature shift. * Design Challenge: The challenge arises in structuring this in JMP as a design factor. If I treat the time of the temperature shift as a continuous factor, there's no straightforward way to represent the 'No Shift' scenario. My initial thought was to create two factors: * A categorical factor indicating whether the shift occurs (Yes/No). * A continuous factor that specifies the time of the shift (1h, 2h, 3h) if it does occur. * Concerns: My concern is about creating illogical combinations, such as a 'No Shift' row with a specified time point. I understand JMP allows certain treatment combinations to be disallowed. Still, I am perplexed about whether this approach is recommended or if alternative strategies are better suited for such a design. Regarding the nested design mentioned by P_Bartell, I'm unsure if it fully applies here, as the 'No Shift' condition is not exactly nested within specific time points. Finally, this is not a degradation-style study; it's more about optimizing conditions for maximum yield. In response to Statman's questions, other factors like nutrient levels might be considered besides temperature shifts, but the focus is currently on temperature. I hope I clarified my question a bit more. Thank you for your support and engagement in this discussion!

nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

How to Include a 'No Shift' Scenario in a Temperature Shift DOE

Nicola_Nuti

I'm currently working on a Design of Experiments (DOE) for a process that involves temperature shifts at three different times. My challenge is incorporating a 'No Shift' scenario as one of the factors.

My initial idea was to:

Treat the three time points for temperature shifts (1 hour, 2 hours, 3 hours) as a continuous factor.
Introduce a 'No Shift' scenario as a categorical factor (yes/no).

However, I'm struggling with how to implement this without creating illogical combinations of 'No Shift' with a time point.

My questions are:

Is there a way to set up a mixed-level factor that combines continuous and categorical levels without creating invalid combinations?
Are there alternative approaches that would be recommended here?

Thank you very much!

Accepted answers

All comments

P_Bartell

I'm definitely not sure of what you mean by shift/no shift from a practical point of view...but in a general sense, within the JMP Custom Design platform there is a way to what JMP calls 'disallow treatment combinations'. You'll end up with some form of an optimal design which JMP will create for you based on the model you've specified, the disallowed combinations, and other design characteristics such as number of runs you specify.

Nicola_Nuti

Hi, thank you for your answer.
What I mean practically is that I will have a temperature shift (from 37C to 35C) in my process happening at different time points, but I would also want to test a treatment in which the temperature shift never happens.
The issue is that if I make "temperature shift time (h)" as a continuous factor, I have no way to express when the shift does not happen at all. So my idea is to create two different factors for this, one that expresses categorically if the shift happens (yes/no) and one continuous that describes when the shift happens (1h, 2h, 3h) if it does happen.

I understand that it is possible to have disallowed treatment combinations in JMP, and I will look into that. However, I would like to know what other strategies might be used to generate such a design and if the one I choose is the recommended one (I can't be the first one to have this problem!).

Thank you very much for your support!

P_Bartell

Here's my thoughts and maybe I'm still not understanding your thinking but it sounds to me almost like a nested design where the 'shift' is nested within the time points. Within each time point you'd have two levels...one with the shift, the second with no shift. Then measure the response for each combination. But that's where I'm confused about the need for disallowed combinations...are you saying all the 'no shift' combinations should NOT be included in the design? This sounds almost like a degradation style study? Like a dissolution problem in pharma?

statman

Quick questions,

1. What questions are you trying to answer in the study?

2. What is your objective? Understand the mechanism or just pick a winner?

3. Can you have shift/no shift at each hour?

4. Are there any other factors?

Nicola_Nuti

Hi everyone,

Thank you for your responses and questions. Let me clarify my experimental setup and objectives to provide a better context:

Experimental Setup: In my bioreactor process, I'm examining the effect of a temperature shift from 37°C to 35°C on antibody production using mammalian cells. The shift can occur at different times: 1 hour, 2 hours, or 3 hours into the process. However, I also want to include a scenario where this temperature shift does not happen.
Objective: The primary purpose is to maximize the yield of the antibody. I aim to understand how the timing of the temperature shift impacts the yield and how the yield compares when there is no temperature shift.
Design Challenge: The challenge arises in structuring this in JMP as a design factor. If I treat the time of the temperature shift as a continuous factor, there's no straightforward way to represent the 'No Shift' scenario. My initial thought was to create two factors:
- A categorical factor indicating whether the shift occurs (Yes/No).
- A continuous factor that specifies the time of the shift (1h, 2h, 3h) if it does occur.
Concerns: My concern is about creating illogical combinations, such as a 'No Shift' row with a specified time point. I understand JMP allows certain treatment combinations to be disallowed. Still, I am perplexed about whether this approach is recommended or if alternative strategies are better suited for such a design. Regarding the nested design mentioned by P_Bartell, I'm unsure if it fully applies here, as the 'No Shift' condition is not exactly nested within specific time points. Finally, this is not a degradation-style study; it's more about optimizing conditions for maximum yield. In response to Statman's questions, other factors like nutrient levels might be considered besides temperature shifts, but the focus is currently on temperature.

I hope I clarified my question a bit more. Thank you for your support and engagement in this discussion!

statman

Thanks for the additional information. I must say this sounds a little OFATish. Why not study all factors in the same design? Agreed, it does not seem nested to me.

I'm still not sure I understand enough, but here are some ideas:

1. One factor (shift) 0 or 2º. At 1 hour intervals take samples and measure the antibodies (you might start at 0 and look for delta antibodies). If you want you can apply another treatment (either 0 or 2º) at those intervals.

2. Two factors: Temp (0, 2º) and Time (1, 2, 3). Factorial gives 6 treatment combinations. Measure antibodies at each hour as well.

Antibodies.jmp

Nicola_Nuti

Hi, thank you for your answer. To clarify further, the antibodies will be measured only at the end of the entire process. Also additional factors will be tested as well; the study is not OFAT. My question is only about this factor as it's difficult for me to formalize it, but there are many other items tested.

As far as I understand, this is what is suggested, am I right?

Temperature shift (yes/no)	Temperature shift time (h)
yes	1
yes	2
yes	3
no	1
no	2
no	3

The issue that I see here is that treatments 4, 5, and 6 are essentially triplicates of the same "no shift". Isn't this wasteful in terms of treatment allocation?