Interaction in 3-level problem

[sarabjeet sidhu]

Hi
I am doing my analysis on ED machining.
i have selected 6 factor i got a results as follows
I have used L27 orthogonal array. Factor A assigned to Col1, Factor B
assign to col2, Factor C to 12, Factor D to 13, factors E to 5 and
factor F to 10

Factors Sum of squares DOF Variance F-ratio Pure Sum % Contribution
A 7005.017 2 3502.508 2.882 4594.642 10.473
B 4107.358 2 2053.679 1.686 1696.983 3.868
C 2714.143 (2) Pooled
D 290 6.939 2 1453.469 1.206 496.564 1.131
E 8156.681 2 4078.34 3.383 5746.306 13.098
F 1550.706 (2) Pooled
Error 21693.372 18 1205.187
Total 43869.369 26

Please tell me is it a feasible results. On studying interaction i
have to reserve 2 columns For eg. for AXB 3,4 col are reserved and so
on and i got a results in ANOVA tables corresponding to each reserve
table.
Please tell me what that two interaction column in ANVOA table
results intricate and on POOLING we have pool both the interaction
results.
Thanks

[Sarah Rashid]

hi, in my work i use S/N ratio i think it is easier than ANOVA

On Dec 16 2011, 12:00 pm, sarabjeet sidhu

[Alois Fabiani-Bello]

the S7N Ratio is of different application of ANOVA analisys... please read mora about this tools.

 

regards.

2012/2/5 Sarah Rashid <sarah....@gmail.com>

--
_______________________________
Prof. Alois C. Fabiani-Bello
Departamento de Ingeniería Industrial
Instituto Tecnológico de Ciudad Juárez

[Ranjit Roy]

Dear Members,
In the recent past, the discussions involved three areas: (1)
Interaction between two 3-level factors, (2) ANOVA, and (3) Analysis
using S/N ratios of results. Although, they are all part of analyses
of DOE results, they each deserved to be addressed individually.

Here are a few brief comments and observation for each.

(1) Interaction between two 3-level factors: To study interactions
between two factors both of which are 3-level factors, you will need
to reserve two 3-level columns (say L-27). Accordingly, when examining
ANOVA, the cumulative effects of the two columns must be consider for
the interaction effects.

As most who are studying interaction realize, while designing
experiments to study interactions, is made easier by availability of
the Taguchi TRIANGULAR TABLE, the evaluation of interaction effects
and determination of the corrective action necessary are much
cumbersome. The process of finding out what to do with interaction for
2-level factors (AxB), that is what level s of A & B to select for
optimum condition, is done relatively simple with two-lines plot of
interaction. This process, however, becomes much complicated when you
are considering interaction between two 3-level factors, as the
interaction plots to consider is of 3 lines each of which has two
segments.

My general suggestion is that, study interactions between 3-level
factors only if you are comfortable with what to to do when you find
out you indeed do have interactions. Otherwise, always study
interactions when factors are at two levels each (force the factors to
be at 2-level if you must study interaction).

INVITATION: If you completed one or more study with interaction
between two 3-level factors, please share with us how you determined
the presence of interaction and how you adjusted the factor levels
based on interaction information.

(2) ANOVA: ANOVA creates insight information about your system just
like an X-ray of your chest provides confirmatory information for the
physician about your body. There are several benefits of ANOVA the
main among which is that it tells you (% influence) in numeric terms,
the RELATIVE INFLUENCE of factors and interactions to the variability
of results. While the factor and interaction information is also
indicated by the slopes of the MAIN EFFECTS plots, the ANOVA is a
preferable and statistically valid information as it also indicates
the collective influences of factors other than those included in the
study (Error term). ANOVA can be performed regardless of your choice
to pursue standard analysis (with single or multiple sample results),
or when performing analysis with Signal-to-Noise (S/N) ratios of
multiple sample results. All rules for POOLING & TEST OF SIGNIFICANCE,
etc. remain the same.

When possible, ANOVA should be included in the complete analysis.

(3) Analysis using S/N ratios of results: When you have results
obtained by testing multiple samples in each of the trial condition,
you should consider performing S/N analysis. And when you perform both
standard and S/N analyses, be aware that the conclusions regarding
effects, influence, optimum, etc. can be slightly different. But, for
better statistically valid performance prediction, you should rely on
S/N analysis.

Beware and keep in mind some basic characteristics of S/N:
- Higher S/N values are desirable no matter the quality
characteristics of the original results.
- S/N values (Main effects & factor contributions/optimum, etc) may
either be negative of positive depending on the quality
characteristics and the magnitude of test results.
- You need to transform results back in the original units of
measurements before presenting to people not familiar with S/N.
- You need to make 6 points increase in S/N to reduce standard
deviation to half its original value (double Cpk value)

- RKR 2/10/2012



On Dec 16 2011, 4:00 am, sarabjeet sidhu

<surinder_sidhu2...@yahoo.com> wrote:
> Hi
> I am doing my analysis on ED machining.

......................