Tutorial wanted for interpretation of Minitab GLM Output

Hi Barbara,
thanks for infos.
I am a beginner to minitab and i have used it for anova.(based on taguchi analysis - orthogonal array).
where can i get a tutorial explaining the conclusions that can be made out from the results of Anova GLM.
Also i want to know the meaning of terms in the results window related to my work and their meaning or implications so that i could draw the conclusions.
for example: if the s value comes this much, what is the effect of that particular factor - like that??
kindly help me with this....

Re: Tutorial wanted for interpretation of GLM Output

I split this off an unrelated thread and moved it into the Minitab sub-forum for better responses.

Re: Tutorial wanted for interpretation of GLM Output

sethu89 said:

where can i get a tutorial explaining the conclusions that can be made out from the results of Anova GLM.
Also i want to know the meaning of terms in the results window related to my work and their meaning or implications so that i could draw the conclusions.
for example: if the s value comes this much, what is the effect of that particular factor - like that??
kindly help me with this....

Click to expand...

If you are new to Minitab, you may not have found its Help features. While not perfect, it is pretty good in many areas. Try navigating to the ANOVA - GLM screen then clicking the HELP button in the lower left. This will bring up a context sensitive help screen on GLM. Look in the top row for a hyperlink called EXAMPLE. This will take you to a GLM example. Read through it and at the bottom, you will find a section called Interpreting Results. It is not extremely detailed, but will get you started.

Note: It may make some assumptions that you already understand ANOVA and p-values. You can navigate to the 1-way ANOVA screen and check the HELP button there and follow that example for more basic information.

Re: Tutorial wanted for interpretation of GLM Output

In addition to the informations in the Minitab help you'll find more in the StatGuide. Just use a right-mouse click to open the context menu in a section of the session window, chooose "StatGuide" (last entry in the context menu) and you'll get more information on analysis and interpretation.

If you have specific questions concerning your analysis and the conclusions for your field of work, please attach the Minitab project file (in a zip-archive) and give us a little bit more details about the issues.

Re: Tutorial wanted for interpretation of GLM Output

Thanks for the info
I tried. but with ended up with an error.
Minitab ANOVA Error Message - *ERROR* No data in column.

what may be the mistakes that i had done??
thanks again....

Re: Tutorial wanted for interpretation of GLM Output

As you see i am in need of analysing the factors(depth of cut, feed, rake angle, cutting speed) which are influencing the responses(temperature, cutting force, thrust force).
I have done 15 experiments with various configurations as listed in excel

when i execute GLM for this i am getting error "no data in column"

kindly suggest me a solution and how can i proceed with this.
thanks again for the reply...

Re: Tutorial wanted for interpretation of GLM Output

First you should add the numeric variables additionally as covariates in a GLM. Assuming all four "factors" are numeric, all should be used in the model field AND as covariates.

Since your design is a Taguchi DoE, you don't have enough informations about possible interactions to estimate their effects. No model can provide reliable estimates for interactions in a design like this.

You can estimate the four main effects (depth of cut, feed, rake angle, cutting speed) and 2 ouf of 3 quadratic effects (depth of cut^2, feed^2 or rake angle^2). There isn't enough information about a quadratic impact due to cutting speed, so this can't be estimated at all.

If all four variables are numeric, you can evaluate the following model or term structure:
'depth of cut' 'feed' 'rake angle' 'cutting speed' 'depth of cut'* 'depth of cut' feed* feed 'rake angle'* 'rake angle'
(in a GLM, a General Regression or a Response Surface Design, doesn't matter which one since the formulas are the same).

Since the residuals are not homoscedastic (=don't have the same variability in the response interval) it is recommended to stabilize the model e.g. using a Box-Cox-transformation for the response (only available in "General Regression" menu) or to be more precise: a different Box-Cox transformation for all of your three responses. The responses should be analyzed seperately, because the model structure is different (=the significant terms are different for the responses).

Hope this helps