What is the plan of experience that I had to choose?As part of a study on a cake recipe, I am trying to find out the effect of an ingredient on the water activity of the finished product (measured parameter which is fixed or belongs to an interval of values). I am not limited by constraints to set the limits of the experimental plan. I wanted to know in this case how I can make an experiment plan.
The objective is to determine the quantity of this ingredient which allows me to obtain a precise value of water activity while having the possibility of varying the other compounds which constitute the recipe.DOERe: Which plan of experiment should I choose?Hello @Syrine ,
In order to better help you, can you give more details on:
- The number of factors and their types (continuous/discrete numeric, categorical, ...)
- Possible dependencies between factors: can the ingredients be varied independently or do some of them have to have a constant total sum for example?
- What is "water activity"? How is it measured? Is this the only answer? You mention a "fixed or within a range" measured parameter, I'm not sure I understand.
- Have you already carried out preliminary experiments with these factors, which could make it possible to create a plan taking this information into account?
- Are you only interested in the effect of this ingredient on the response? Do the other ingredients in the recipe necessarily have to be varied in their levels?
Here are some questions that will help you see more clearly and help you more effectively.Re: Which plan of experiment should I choose?Thank you for your reply,
In fact the variables are numerical variables (amount in grams). In fact there are variables which are dependent and others independent. In reality I wanted to substitute an ingredient in the reference recipe with another element but by testing several I will vary the quantity of this substitute until I have the same effect as that in the reference recipe. This compound has an effect on the water activity (the available or free water in the food and which could cause the growth of micro-organisms). The substitute has a sweetening power is therefore by increasing its quantity, we should decrease a quantity of sugar, it is here that the dependence between the variables is presented, otherwise for the other variables, they are independent.
For the water activity response, we aim for a value but we can say that it is an interval.
The measurement is made using a measuring device
I did not do any experiments beforehand
And thank youRe: Which experimental plan should I choose?Good evening @Syrine ,
It remains difficult to help you without more information on the system to be studied.
Is it possible to describe it more, even if it means anonymizing/coding the name of the ingredients? Specify how many and which are dependent on each other (eg ingredients 1, 2 and 3)?
- How many mixing variables (variables dependent on each other whose total quantity is fixed) do you have?
- How many independent variables do you have? For each, what is its type (continuous, categorical, etc.), its range of values (high and low for continuous numerical factors, or the number of levels for categorical factors)?
- Is the ingredient to be substituted part of these mixing variables or is it an additional independent variable (whose quantity can be freely chosen without impacting the quantity of other ingredients)?
- Are there several possible substitutes for this ingredient that you wish to test (alternative A, B, C, ... : categorical factor with a number of levels/alternatives?) ?
- Do you have an experimental budget (number of experiments) limit that you would like not to exceed?
Since your plan seems to mix mixing factors with continuous factors, it seems appropriate to use the "Optimal Plans" platform, as it will give you more freedom to define and mix different types of factors and choose the model and the terms to be included, as well as some flexibility in choosing the number of experiments.
For the subject of the target response value, you will anyway have in the analysis of your plan water activity predictions, with their confidence interval. Depending on the type of plan, the variability of the response, the number of experiments, the level of confidence (95% for example), ... this interval will be more or less large, and you will thus be able to verify that the predicted optimum value (and its confidence interval) for a certain combination of values of your factors corresponds to your target and that the deviation from the confidence interval is sufficient for your needs.
Edit: If you want more information or training on Design of Experiments or if this topic is your first Design of Experiments, I highly recommend the following free JMP trainings:
- Introduction to design of experiments: Design of Experiments Intro Kit | Getting Started with JMP
- Statistical Thinking for Industrial Problem Solving (a module is dedicated to DoE): Statistical Thinking (STIPS) - Free Online Statistics Course | JMP
- Other resources or examples: Design of Experiments | JMP
This content can allow you to build your skills on experimental plans, as well as to be able to describe more precisely the plan you want to carry out.
Re: Which plan of experiment should I choose?Hello,
There are also simulation software that use databases of hundreds of ingredients, additives, .. and which will calculate the aw.
= another search strategy.
cordially
Re: Which experimental plan should I choose?Thanks for your reply,
Can you please send me the link of simulation software? Maybe that could help me.
Best regardsRe: Which experimental plan should I choose?Hello @Francoise and @Syrine ,
Indeed, if models already exist, this approach can be complementary to the experimental design, by building an experimental design adapted to the problem, and using values resulting from these simulations as an answer (having checked beforehand the scope and the hypotheses of these models if possible, for example checking whether the models/predictions take into account mixtures with a variable number of compounds, or only binary/tertiary mixtures).
This combination of techniques (simulations and experimental plan) can make it possible to quickly identify a formulation area of interest (by carrying out the minimum number of simulations for a maximum gain in information on the experimental field, rather than a "random" search by iterations), and to validate its relevance and robustness by a few experiments.
While doing some research, I came across the Aw Designer software (paid).
@Francoise , did the option I gave you for your ellipse topic on the Community solve your problem: Re: graph with ellipse - JMP User Community
Re: Quel est le plan d'experience que je devais choisir ?bjr,
il y a eu bcp de modélisation faite, surtout dans l'application de la microbiologie prévisionnelle., c à d pour voir l'impact de l'aw sur le développement des bactéries, en particulier les pathogènes.
la première étape du plan d'expérience passe par le recueil de l'état de l'art et des connaissances, donc par une bibliographie entre autre. Aujourd'hui, avec internet, vous pouvez trouver bcp d'informations (articles scientifiques, livres, présentations pdf ou pwp,...)
cdlt
Re: Which experimental plan should I choose?Hello
If I understood correctly what you mean, it is that with the aw software I can eliminate several experiments resulting from plan of experiment. But what I didn't understand is how to verify the model's prediction.
I've never done an experiment plan other than the central composite plan so I don't know too much about the realization of the other plans.
Best regardsRe: Which plan of experiment should I choose?Hello
I did research on the ingredients to use in the recipe but not on the prediction of aw, because it is according to the ingredients that I will put, it can change.Re: Which experimental plan should I choose?Hello @Syrine ,
I came back to Françoise's proposal to use simulation software in order to be able to directly predict the water activity results in relation to your mixtures, and thus quickly find the optimal formulation zone. From the presentation shared earlier ( Aw Designer water activity calculation software (slideshare.net) ), the software is able to take into account and calculate the water activity for a wide variety of mixtures (see slide 21).
What I proposed is a combination of the two approaches (design of experiments and modelling) in a few steps:
- Create a design of experiments suitable for your experimental space (for this, see my questions, answers and resources about the designs of experiments mentioned above: Re: Which experimental plan should I choose? - JMP User Community ),
- Use simulation software adapted to your needs (I do not know Aw Designer personally, so no idea if it is adapted to your needs (and at your expense!)) to fill in the answers of the design of experiments with the results of modeling of this software,
- Analyze the experimental plan, find the optimal formulation zone corresponding to your need/target/range of values,
- Carry out validation experiments with the factors having their concentrations/levels fixed in the optimal zone found previously during the analysis (to verify that the results obtained by the simulation software are indeed consistent with your needs).
The combination of these approaches can save you some manipulation time: instead of doing all the experiments required by the plan + the validation points of the model that you will have created, you would only have to do the validation points, the other experiments being modeled using appropriate software.
Validation points, made experimentally, allow you to verify the model's prediction (and the results of the simulation software if you find significant value deviations).
If you have little experience with design of experiments, you can consult the resources mentioned above, and detail a little more what you want to do (see questions) so that members of the community can help you.
Hoping this answer will help you see things a little more clearly,
