Hello,
I'have got a history of 22 samples that were tested and i've got 2 samples that doesn't have a default. So 91% have a défault.
In order to be able to make an experimental plan i need to reduce my number of samples by test (using a tagushi table) and be sure to catch my samples with default.
I need help with a mathematical method to prove how i chose a low number of samples.
Thank you!!
HI Larbi,
You may need to state what you are trying to prove or show (Hypothesis)
Based on the data you have at the moment you have 91% with a default.
The 95% CI for the true number with a default ranges between 79.04% to 102.96%(100%)
Andrew
HI,
You can try statistical methods to select influencing parameters (Anova Test, Regression model,...)
Otmane
Larbi, first of all, as Otmane mentioned, you need to determine the correct varaibles (A, B, V, etc). Then you have to decide the accuracy (reliability R) on which you want to perform your study. From provided historical data your R index is R=0.91 (this implies a failure probability of F=1-0.91). Second, you have to choice the suitable Taguchi orthogonal array (OA) (suppose a L9(3^4) was selected. And finally, for each row of the selected OA (the L9(3^4) has nine rows) you has to test (measure) n=-1/ln(R) items (n=-1/ln(0.91)=11parts. It is to say in each row (setting) of the selected OA you has to test at least 11 parts in order to have as minimum of 91% of probability to observe the desired response. Finally observe that: 1) this 91% cannot be increased without introduce in your study a form to increase the severity on which, historically the selected variables (A, B, V, etc) had been affecting.[Formula n=-1/ln(R) is given as Eq.(45) in my paper "Weibull and lognormal Taguchi analysis using multiple linear regression"].
2) If you deside testing lower n2 parts (n2
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis