Minitab 16 - Science topic

Greetings there,

To be laconic, I think if u can go through it, it might help.

is that a confusion matrix? very untidy. Specificity and sensitivity you can calculate the following way:

However you can do both ways i.e. individually and over all the TP's and TN's, FP's and FN's. However, looking at it class wise would make much more sense.

Sensitivity: TP/(TP+FN)

Specifity: TN/(TN+FP)

You can used RSM in three factor, but I will advice you to tanguchi or full factorial b/c RSM you mighty not get accurate result, is more better when you applied it in Four factors five levels or more

In order to study the interaction between three parameters,  you need to fit your data with cubic regression model. However, standard response surface methodology such as central composite design can not fit cubic regression model. You can study the interaction between three parameters using cubic regression model by using custom design of experiments such as D-optimal design or I-optimality criteria. 

You could transform this equation by taking the logarithm of both sides of the equation to convert it to a linear equation and then perform linear regression. After that you can the anti-log.

Here is how your equation will look like after the logarithmic transformation:

Log(M) = x.log(A) + y.log(B) + z.log(C)

Hope this helps answer your question.

Hello. 1- If you are refering to the Pareto chart for the standarized effects in a design of experiment, the reference line is the quantile in the Student's t-distribution or the Lenth’s pseudo-standard error (PSE) . See e.g. http://support.minitab.com/en-us/minitab/17/topic-library/modeling-statistics/doe/factorial-design-plots/what-is-a-pareto-chart-of-effects/ 

For the Lenth’s Pseudo-Standard Error see http://www.tandfonline.com/doi/abs/10.1080/00401706.1989.10488595#.VsYoquYposI

2- If you mean the traditional Pareto chart the dotted line is the cumulative frequency. See the description here (http://support.minitab.com/en-us/minitab/17/topic-library/quality-tools/quality-tools/pareto-chart-basics/).

Regards

Sorry...I didn't read the whole question. In the case of F statistic or p-value, it means the value could not be calculated. Sometimes there are not enough degrees of freedom, Missing p-values and F-statistics will occur in an ANOVA table whenever you have a 2-level design with one replicate, and you include all the terms in your model.

i hope that helps. 

Hi

Its very simple to do ANOVA analysis in Minitab

Open Minitab, go to "Stat" then to "ANOVA"and to "General regression"

Chose the response and the model type then you will have ANOVAanalysis and 

Hi KS,

Thanks for the article, let's look at some facts here. The initial production line is unbalanced, the longest cycle time is 47 seconds and the average output calculated is 62 seconds. This is due to the asynchronous process creating idle time and waste. Would be better to have station close to each other with no WIP in between and running at about 50 seconds automated takt, then grouping station 1 and 2 then 7 and 8.

But that was not your question. Normal distribution supposes that your variable can vary freely from minus infinite to plus infinite, and time data rarely meet this requirement, at least in real world. I have almost never seen real operator time data normally distributed as the probability to expand in plus direction is large and usually there is a hard limit in minus direction. Then why having time from operators with 3 digits? Of course it helps finding it normally distributed but time with operators recorded with 1 second resolution is enough, then by analyzing the distribution with 1 second classes we can see the modal value which is the most frequent one with the higher probability of occurrence, more than the average value.

I do not see in the article some foundation for the Mann-Whitney test. They talk about average and then do a test about median. With the 6 run's average a simple IMR chart shows that anywhere between 3243 and 3252 is the same thing.

Now if I create a simple calculation method, the simulation, that calculates 3247 with moderate variance, and I do a test to see if I find 3247 ???? What is the point? Do I demonstrate that the calculation repeat itself or that the simulation works?

Simulation is nothing more than denying reality of variance of some factors, when it is not ignoring them totally. Looks like the purpose here is to demonstrate that the software is good and we must buy it, I will not.

Best regards, Luc

its a static problem, and you may use smaller is better as we desire a lower coefficient of friction and  frictional force

You are not clear about your design,

So, I assume you are doing one way ANOVA:

Try this path:

Stat / ANOVA / One way ANOVA: put the codes of your treatments in 'factor" box and the response in "response"; click "comparison button", write the error rate: 5 for 95 and 1 for 99 percent confidence.

click ok twice and in results you have:

Treatment   N    Mean      StDev     95% CI

1                    3    5.000     1.000     (2.742, 7.258)

2                    3    7.33       2.08        ( 5.07, 9.59)

3                    3    7.667     1.528      (5.408, 9.925)

the data in 95% CI column are critical values.

Regards,

Vinod -

I am not clear on your context here, but in regression, the points on the regression line or curve are the predicted values, say y*, and the observed values are y. I factor the residuals into a random factor, and a nonrandom factor based on the regression weight, the product of which we will call e, such that we have y_i = y*_i + e_i . 

So the answer to your question depends upon your goal. If you are looking for an observed value, you take that which was observed. But perhaps you are more interested in the model?

I do not know if what you refer to by 'experiment' is a single observation, or a set of observations from which you estimate/'calculate' one aggregate observation, and whether the predictions are from one model or different ones, the details of which might be important, but it would seem that regardless, the answer to your question depends upon whether you want to know what was observed/calculated, or modeled. If you are looking to future use, the models may be more important. Regardless, you should also consider estimating the variance of the prediction error.

Cheers - Jim

Why are you using MANCOVA? What is your covariate? 

If you go to Stat>ANOVA>General Manova in Minitab, you can add your factors of interest in the "Model" box, your responses in the "Responses" box. There is also an option for adding covariates. 

Dear Sri Abu

Great Question. Historical std. dev values are based on the info about a process in a particular environment... For example in an industry that mastered the art of drilling, their std dev is from the past data... however it may be different in a process of milling within the Industry... However, in the absence of historical info, it is Industry practise to assume 3 standard deviation on the tolerance of the spec.

Hope it helps... if you have specific question on R&R, pl do feel free to post it to me thro my home page.

Kindest Regards

Ramanan

Dear Deewajar

If you add parameters to a model its Rsq always grows. This is the reason behind the adjusted Rsq. You can use Rsq to compare linear models of the same number of parameters, but if you have two models of different number of parameters it is best to use other measures like the adjusted Rsq or the Cp of Mallows.