# Correlation regression for transactional

Six Sigma – iSixSigma Forums Old Forums General Correlation regression for transactional

Viewing 17 posts - 1 through 17 (of 17 total)
• Author
Posts
• #36505

Sorour
Participant

can anyone help I have been tasked with teaching correlation regression next week for transactional green belts has anyone any suggestions for a class activity prefreably without the use of mini tab?

0
#105474

Mikel
Member

Excel has all you will need under Data Analysis

0
#105477

Sorour
Participant

Thanks for this Stan have you any ideas for an in class excersise?

0
#105484

Bob J
Participant

Paul,
I’ve used the student’s height and shoe size for a correlation exercise before with some success…  It’s reasonably fun and off the wall enough to be interesting without having an obvious result.
Hope this helps…
Best Regards,
Bob J

0
#105487

Sorour
Participant

Thanks Bob great example and a good class excersise, where I am struggling is trying to find a use for the tool in a transactional process. Any thoughts?

0
#105489

Bob J
Participant

Paul,
No problem…  Regression (particularly Binary Logistic Regression) is a great tool for transactional processes.  We’ve had projects that have used it to improve exworks OTD, Quotation hit rate, Bid project profitiability, product reliability in specific environments and Call center customer sat.
Hope this helps…
Best Regards,
Bob J

0
#105491

Sorour
Participant

Thanks Bob
it’s back to Breyfogle for me, half way through BB training and binary logistic regression is over my head at the moment.
Thanks for the examples though at least I know where to start looking, nothing like being thrown in at the deep end.
Cheers
Paul.

0
#105547

Bob J
Participant

Paul,
My pleasure…  Glad to help…
Best Regards,
Bob J

0
#105553

KBailey
Participant

Instead of Excel’s Data Analysis toolpack, you might want to just do a line chart. On the Options tab of the Add Trendline window you can select to display the equation and R-squared.
Height/shoe size is a good example to teach how, but some people still won’t see how to apply the concept to transactional processes. Think of the key metrics you would apply to transactional processes: time, touches, dollars, defects, success rate, reject rate, etc.

0
#105586

John J. McDonough
Participant

binary logistic regression is over my head at the moment
Don’t let it spook you.  These statisticians just like to use big words to impress people.  There’s no magic, and it is pretty cool.
All it means is that with a little mathematical hand waving, you can do regression on categories instead of centimeters.  Quite often that’s what you have in transactional processes.
–McD

0
#105593

Sorour
Participant

Thanks John
A calming influence! have you any suggestions fo examples I could use in training?
Cheers
Paul

0
#105594

Sorour
Participant

Thanks K
Right on the nail! I need to some how relate a class excersise (trying to make it fun) to actual work based data. do you know if the height & shoe size produces a positive correlation? really need to draw out prediction equation from the excersise.
Many thanks for the input all help greatfuly accepted.

0
#105602

John J. McDonough
Participant

Paul
Lo those many years ago when I took Black Belt training, we had some examples that weren’t all that engaging.  For binary logistic regression we had a prediction of whether a day would be rainy or not.  For ordinal logistic regression we had sky cover numbers that were categorical, like 1, 2, or 3, and developed a model based on humidity or some such.
As I recall, the multiple predictor example was more fun.  Seems like we had data on the Titanic passengers and developed a model on whether they would survive based on the class they were traveling in, and maybe their ages and sex or something like that (was a while back).  It seems like there were quite a few passengers, though, so it was a good sized data set.  I have no idea where you could find the data.
In real life I’ve seen logistic regression used to predict defects in some software development deliverables.  Also seems like we used it somewhere in the I/T infrastructure to predict some sort of failure but I’m afraid the old gray cells aren’t giving up the details.  More recently I’ve been using regression to understand where we are headed with amateur radio public service activities and to try to fish out what affects people’s participation.
For ordinary regression, most of our examples were from manufacturing, but we did have one where we tried to model the time to pay as a function of invoice amount.  (If I recall, the correlation wasn’t significant based on the data we were given).
Another model we did, I think it was a multiple regression model, was on salary.  I think we had things like age at hire, length of service, sex, etc.  The idea was to show whether there was a bias based on sex.  I don’t recall all the variables we had but it seems like there were plenty of columns that made it more interesting fishing out the relevant ones.
This is all pretty fuzzy and you may have some trouble locating data for these specific examples, but maybe it will get your juices flowing.
–McD

0
#105616

Sorour
Participant

Thanks John really appriciate your help.

0
#106035

Packrat
Participant

I’d suggest you ask them to correlate SS# with height, weight or anything else.  The sofware will give you a correlation but it does not mean anything. The teaching pont is that you have to know what you are correlating or you will report a meaningless result.

Packrat

0
#106151

Ram
Participant

Paul,
Did you try K’s ideas of using Excel? How was the class?  Did you survive? Or you trhrow big words like Binary Logistic Regression?
If you are still interested please share a transactional project example from you or from your GBs
Ram

0
#108957

kate
Participant

Hi Bob

Can you please provide me the example of Regression analysis that u did to improve callcenter csat . My email is [email protected]
Appreciate the help
Thanks

0
Viewing 17 posts - 1 through 17 (of 17 total)

The forum ‘General’ is closed to new topics and replies.