How should I go ahead with my experiments (interpretation) when I know before hand that my response (continuous) will have a time delay.
How do I incorporate time delay in interpreting the results of my experiments?
Thanks,
Ashish
@BivariateKoala1 wrote:
How should I go ahead with my experiments (interpretation) when I know before hand that my response (continuous) will have a time delay.
How do I incorporate time delay in interpreting the results of my experiments?
Thanks,
Ashish
Hello,
Here's how to consider time delay when interpreting the results of your experiments with continuous responses:
Understanding Time Delay:
Definition: Time delay refers to the lag between when you make a change to the system (independent variable) and when you observe the full effect on the response variable.
Impact: This delay can complicate interpreting results because the observed response at a specific time might reflect a previous change in the system.
Incorporating Time Lag in DoE (Design of Experiments):
Identify and Quantify Time Delay:
Conduct preliminary tests to estimate the time delay between changes and their full effect.
Adjust Data Collection:
Shift data collection points to capture responses after the expected time delay.
Consider collecting data at multiple points after the change for a more complete picture.
Model Selection:
Use statistical models that account for time delays, such as:
Autoregressive Integrated Moving Average (ARIMA) models
Transfer Function models
Data Analysis:
Analyze the data considering the time delay:
Focus on responses after the expected lag to avoid misinterpreting the impact of changes.
Use appropriate statistical methods for time series data analysis.
I hope the information may help you.
Hi @BivariateKoala1,
Welcome in the Community !
I'm not sure to understand the situation about the time delay :
- Is the time delay for the response the same, no matter the experimental combinations of factors ? Or is it dependent on the combination of factors ?
- Can you record the response from initial time T0 to the time where you have a change in the response ?
Depending on your answers, here are some suggestions :
- You could create a time-robust DoE, by taking into account the possible effect of time/order in the design. Here is a similar topic where this type of DoE has been used : Covariates in defined order in custom design
With this kind of design, you balance the levels of factors accross time, to avoid confounding/aliasing between time or experiments order and any of the studied factors.
- If you can record the response across time, you could maybe transform the response into a time-independent responses, or use the functional response and analyze it through different platforms, like Functional Data Explorer in JMP Pro to directly create a functional model for your response, or using Fit Curve/Fit Y by X platforms to extract the interesting coefficients of your curves and analyze how the change of your factors impact these coefficients.
You might be interested in closely related topics : DoE design for a chemical reaction with time point data
How to obtain the optimum formulation
A combination of these two techniques (if possible) could help you mitigate the effects of time in your design, as well as improving the understanding of the link between your functional response and your factors.
Hope this first discussion starter might help you,
