Four variables (Lt1, Lt2; St1; St2) are influencing my data (modulus value) points.
A) I would like to evaluate the effect of each variable on the modulus value while the other three remain constant. E.g., is Lt1 affecting the modulus value significantly or only little impact? is there any relation between them?
B) Again, the modulus value, E is a function of four constants/parameters, namely delta (representing min value), alpha (=max value), beta (=position) and gama (=deviation). Now AGAIN I want to evaluate the effect of each variable on these constants/parameters while the other three variables remain constant. E.g., is St1 affecting alpha, beta, gama or delta values significantly or only little impact? any relation between the variables and these constants/parameters?
Any advice is appreciated! Thanks.....
It appears that you would want to start with multiple linear regression
I would like to consider the Taguchi method and ANOVA. Taguchi design of experiments can determine how the individual parameter can affect the final output and ANOVA is useful to evaluate which parameter is contributing how much.
Hi Dalim,
You can do three steps as follows:
a) % Error Analysis
b)Perturbed Analysis
c) Uncertainty Analysis using Monte Carlo
% Error Analysis:
E=f(A,B,C,D). You have base value either from experiments or co-relations. Using them calculate the value of E and then do the same calculation for your model under identical interval like 1% axial strain interval. Then calculate % Error =Abs[(E_Base-E_Model)/E_Base]*100
Here Abs mean magnitude rather than sign.
Perturbed Analysis
Perturbed your parameter with respect to reference value and calculate the men Error. There might have increasing order or increasing to decreasing or decreasing. Which will represent the sensitivity of your parameters on the calculation of E.
Uncertainty Analysis:
For in put parameters considering one to 3 or as you required standard deviation using Monte Carlos Simulation Normal Distribution cal be plotted. Any magnitude t he below the Bell Curve is safe for your model.Here Kurtosis represents the 'peakedness' or 'flatness' ofthe probability distribution.
Regards,
Mohammad
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis