topic Re: Significance of few factors among several in Discussions

Significance of few factors among several

gam1 — Mon, 15 Jan 2024 10:51:33 GMT

Hello everyone,

I would like to study an impact of the six factors on evaporation (at the end I would like to have a “rigorous” model with possible interaction effects and power terms):

I already know that the first four have a strong impact on evaporation and would like to know whether the last two (plate load, volume/well) should also be included in the study (model).

How do you suggest to to set up an experiment? Would it be most efficient to first find out whether the two factors (plate load, volume/well) are significant (how to set up a DOE for this purpose?) and then proceed with DOE (augmented DOE?) to get more accurate information about the factors and their (possible) interactions?

Thank you in advance!

Best regards

Re: Significance of few factors among several

P_Bartell — Mon, 15 Jan 2024 12:31:26 GMT

First question I have is do you really want the Role of all but one of the factors as discrete numeric?

Re: Significance of few factors among several

gam1 — Mon, 15 Jan 2024 14:01:56 GMT

No, it is not mandatory. The factors could also be treated as continuous numeric.

Re: Significance of few factors among several

P_Bartell — Mon, 15 Jan 2024 15:19:19 GMT

Next questions...are you in screening mode? Or optimization mode wrt to your study? What are your specific experimental goals and objectives?

Re: Significance of few factors among several

statman — Mon, 15 Jan 2024 16:02:49 GMT

I do not understand the situation enough to provide specific advice and perhaps my thoughts are not useful.

I do have some questions/comments:

Not sure I understand what the response is? Is it rate of evaporation? Evaporation of what (I assume some kind of liquid)? How is it determined (e.g., delta volume, delta weight...)? Have you studied the measurement process?

"I would like to study an impact of the six factors on evaporation (at the end I would like to have a “rigorous” model with possible interaction effects and power terms):"

Do you want a model for explanation and/or for prediction? (What Pete is asking). The model should be useful and appropriate. Having a complex (rigorous) model does not make it better (I would say the exact opposite is true).

"I already know that the first four have a strong impact on evaporation and would like to know whether the last two (plate load, volume/well) should also be included in the study (model)." How did you study the first 4 factors (e.g., observation, experimentation)?. What levels were plate load and volume at when those 4 were studied? If they were constant, the effects of the 4 factors you studied are contingent on those levels (they are part of the inference space). If they varied, they would contribute to the experimental error. Realize the design space and inference space will impact how useful the experiment will be for prediction.

“Unfortunately, future experiments (future trials, tomorrow’s production) will be affected by environmental conditions (temperature, materials, people) different from those that affect this experiment…It is only by knowledge of the subject matter, possibly aided by further experiments to cover a wider range of conditions, that one may decide, with a risk of being wrong, whether the environmental conditions of the future will be near enough the same as those of today to permit use of results in hand.”

Dr. Deming

Scientifically, it seems likely that higher T (if it is temperature), higher exposure Time, increased shaking speed and no lid would increase evaporation rate. What is the surface area of the material that is evaporating? What are the ambient conditions (e.g., dew point)? How representative is the material that is evaporating during the experiments of future material?

I'm not sure I understand your level setting. Analysis is easier if the levels are balanced and equidistant (e.g., T at 20, 27, 34; ExpTime at 4, 7, 10)

Re: Significance of few factors among several

gam1 — Tue, 16 Jan 2024 06:36:50 GMT

I didn't do any measurements yet. Firstly, I would like to identify whether the plate load and the volume/well are significant factors. Secondly, and this is the main goal of the study, I would like to have a prediction model for the evaporation (rate). The evaporation is unwanted side effect that I would like to reduce.

Re: Significance of few factors among several

gam1 — Tue, 16 Jan 2024 09:20:50 GMT

"Not sure I understand what the response is? Is it rate of evaporation? Evaporation of what (I assume some kind of liquid)? How is it determined (e.g., delta volume, delta weight...)? Have you studied the measurement process?"

The response is a rate of evaporation of a liquid and it is undesirable effect that I would like to minimize. The evaporation rate will be measured by weighting the remaining liquid.

"Do you want a model for explanation and/or for prediction? (What Pete is asking). The model should be useful and appropriate. Having a complex (rigorous) model does not make it better (I would say the exact opposite is true)."

As already explained in response to Pete's question, firstly, I would like to better understand what are the factors that effect evaporation (are the plate load and the volume/well significant factors?) and, secondly, obtain a prediction model which I will use to reduce evaporation.

"How did you study the first 4 factors (e.g., observation, experimentation)?"

I know that the first 4 factors impact evaporation from observation. This is also in agreement with my knowledge of chemistry/physics. I have a "feeling" that the last 2 factors might also have and impact on evaporation. I would like to confirm/reject this by conducting experiments/measurements. I assume that these measurements can be also used later in DOE (the aim of this DOE would be construction of a prediction model).

Re: Significance of few factors among several

P_Bartell — Tue, 16 Jan 2024 13:47:24 GMT

I'm generally a fan of sequential problem solving using DOE. Start small. Learn as you go. Make future decisions based on incorporating new knowledge and information. So starting with a simple 2^2 full factorial in plate load (note I'm recommending only 2 levels for plate load, you don't need 3 to assess significance) and volume/well should give you an idea of their impact. Then depending on how that turns out, plan a second experiment based on what you learn. Too many divergent pathways to give you a single recommendation. Keep in mind noises as you go along in time. Blocking is your friend.

Re: Significance of few factors among several

statman — Tue, 16 Jan 2024 17:59:13 GMT

Thank you for the explanation. While I agree with sequential investigation, IMHO, it is better to start with a large design space and hopefully interpolate. By this I mean lots of factors at bold levels to start. The issue is inference space. The smaller the inference space, the less likely the results will hold true in the future (when this conditions invariably change). If you only experiment on 2 factors, what are the other factors "doing" during this experiment (e.g., are they changing?, are they constant?). Since you have not quantified nor rank ordered the effect of the 4 factors, why not run a lower resolution 6 factor design with the factors at 2 levels (e.g., 2^6-2 res IV, 16 treatments or 2^6-3 res III, 8 treatments). The next iterations can help to understand more complex model terms (e.g., non-linear). At there same time, you might want to run repeats and estimate the factor effects on both the mean and variation of evaporation rate(or at least minimize measurement error).

In the end, I always suggest you design multiple options (no one knows the "best" design á priori. For each option, predict what you can learn (e.g., what effects can be estimated, which may be confounded and which are not in the study). Weigh this knowledge against the resource requirements. Choose one and prepare to iterate. The purpose of the first experiment is to design a better experiment.

Re: Significance of few factors among several

gam1 — Wed, 17 Jan 2024 07:26:47 GMT

I like the idea of "starting small" :) But on the other hand I share the statman's concern:

"The smaller the inference space, the less likely the results will hold true in the future (when this conditions invariably change). If you only experiment on 2 factors, what are the other factors "doing" during this experiment (e.g., are they changing?, are they constant?)."

I guess that in case if there is no significant impact of plate load and volume/well after conducting the 2^2 factorial there is still quite significant chance that the 2 factors still have an impact at some different combination of the 4 (significant) factor that were held constant at the 2^2 factorial experiments.

Re: Significance of few factors among several

gam1 — Wed, 17 Jan 2024 07:50:03 GMT

I don't understand what do you mean with the "hopefully interpolate". Could you explain?

Re: Significance of few factors among several

P_Bartell — Wed, 17 Jan 2024 15:38:17 GMT

In my 2^2 full factorial recommendation, I was reacting to your comment, '...firstly...'. I have no quarrel or issue with designing a larger experiment if you feel there is '...still a quite significant chance that the 2 factors still have an impact at some different combination of the 4...'. You've just described in words a desire to estimate interaction effects among the full process space. So building an optimal design that specifically provides estimation of desired interaction effects is the way to go.

Re: Significance of few factors among several

statman — Wed, 17 Jan 2024 16:02:21 GMT

What is preferred is the design space is large enough to "contain" the optimum conditions (given current materials and technologies). The you can quickly converge on that optimum. If your design space is too small, then you will have to move the current space to achieve the optimum (essentially extrapolate). I will admit, one of the commonest of "errors" in experimental design is to have too narrow of level setting in initial designs. It seems there is some overt or underlying attempt to find optimum in the first experiment (of course, driven by managements' desire for the one-shot solution).

The reasons we are always searching for optimum and never seem to get there is:

1. We never start our investigations with ALL (we always choose some subset), and

2. New materials and technologies are constantly being invented expanding the dimensional space of investigation