Name: Six Sigma Basics: Correlation & Regression
Price: 1007 INR
Availability: InStock
Rating: 4.0 (4 reviews)

Live online Basic

Six Sigma Basics: Correlation & Regression

4(4)

2157 views

COMPLETED

Tell us and we’ll notify you when the next batch is scheduled.

2157 views

Prasenjit Guru

7-day money-back guarantee
Session recordings included
Certificate of completion

Volume pricing for groups of 5+

Why enroll

- You will learn to establish Y= f(x1, x2,...) association between input and output variables. This topic will help participants understanding and optimizing of product or process parameters and predicting output.

Your instructor

Prasenjit Guru

Project Manager

15+ yrs experience

5Courses

4Avg rating

Is this course for you?

You should take this if

You prefer live, instructor-led training with Q&A

You should skip if

You need fully self-paced, on-demand content

Training details

This is a live course that has a scheduled start date.

COMPLETED

Questions and Answers

Q: You're reviewing a supplier MSA pack attached to a control plan. The regression output is pasted as an image with axes unlabeled, and the instrument list shows the torque transducer range as 0–50 N·m. The long-tail phrase is: "regression plot unlabeled axes MSA torque transducer". Based on the scatter and the stated range, what inconsistency should make you stop the pre-startup review?

A: 50 N·m is the hard boundary. A slope that small across the full span hints the x-axis is scaled in a smaller unit, a classic vendor paste error when lbf·in data gets relabeled as SI. Physics doesn't force zero intercept once bias exists, R² thresholds aren't mandated, and heteroscedasticity needs context before killing a startup.

Q: A Black Belt cites ISO 13053 while arguing that correlation alone justifies adjusting a control limit before SAT. The exact phrase engineers search is: "ISO 13053 correlation vs regression control limits". What application of the standard should you push back with?

A: Cause-and-effect is the boundary. ISO 13053 frames improvement around CTQs and verified relationships, not just numerical association. There’s no universal r threshold, scope exclusions like that don’t exist, and p-values alone don’t show causality.

Q: During a pre-startup safety review, you see a simple linear regression used to predict leak rate from pressure. Sample size is 8. The long-tail phrase is: "minimum sample size regression process startup decision". Without software, what's the most defensible stance on the trend reliability?

A: n=8 is the number that bites. At that size, leverage is high and slope stability is poor, making it unsuitable for decisions tied to safety. R² and range help but don't fix sensitivity to single points.

Q: You're checking a control chart where a vendor claims improving correlation between temperature and yield over three trials. The phrase is: "correlation trending wrong direction last three tests". The last three points pass spec but trend opposite to expectation. How should that influence your decision?

A: Three points is the trigger engineers watch. A short run trending the wrong way suggests process shift not yet breaching limits. Specs alone lag behavior, while knee‑jerk shutdowns or cherry-picking data miss the signal.