# Non-Normal Data

Six Sigma – iSixSigma Forums Old Forums General Non-Normal Data

Viewing 8 posts - 1 through 8 (of 8 total)
• Author
Posts
• #48506

Carlos Palma
Participant

Hello:
I need help from proffesional people with experience in Six Sigma. My question is that the data of my process is non-normal, is a distribution plot lognormal, but data is continuos, for this, the hypothesis testing that I have to use, is regression, is this correct? or do I have to use another testing for lognormal data?
Thank you very much for the answers, I need it.

0
#163850

Erik L
Participant

Carlos,
There are a lot of tools that have assumptions wrapped into the data that is used for the analysis.  Regression is one of the tools that we have that doesn’t place assumptions around the distribution that models the core data for either the predictor or response.  Assumptions around the structure of data are applied after fitting the model and looking at what is leftover, unexplained, by the equation.  Hope that helps…
Regards,
Erik

0
#164027

Carlos Palma
Participant

Ok, the solution I have implemented is to transform the non-normal data into normal through the Cox-Box Transformation, by minitab, wirh this Six Sigma tool I have got Normal data, because I have tested the Normallity Test and It was correct. With Normal data I have tested the Regression Test.
Mu question is, Are you ok with this work?, Is correct? Can I trust it?
Thank you very much for your help, I need it.

0
#164028

Ken Feldman
Participant

Carlos, not sure from your post what your research question is.  All you said was that you wanted to do a hypothesis test and were doing regression.  Those statements alone don’t provide information.  Transforming the data is usually a last resort and until more is known about what you are trying to do with the data then transforming might not be needed.  As was stated in an earlier response, regression doesn’t have an assumption of normality for the dependent and independent variables, only for the residuals.  While t tests and anova have assumptions of normality, they are usually robust to it and if not, nonparametric tests might serve you better than transforming the data.  Provide a little more detail and you might actually get a useful answer.

0
#164029

annon
Participant

If you have transformed the data for the purpose of running the regression, this was not necessary.  Assumption of normality is required only during the residual analysis to test for ´goodness of fit´.
Go into MTB ´HELP´ sections and study the examples.  good luck.

0
#164030

RC
Participant

Sounds to me like you are not a Black Belt or trained in Lean or Six Sigma. If you are asking someone to do the data analysis for you the just ask – do not beat around the bush.
I have seen a lot of links on this web site in which it appears that people that are not trained want to have others provide the answers to their projects.
There a a number of non-parametric tests that can be applied to non-normal data if needed.

0
#164031

Regression
Participant

If you can validate the four basis assumption of regression through the transformation, then you can use the equaltion with nor problem. However, all summary statistics Rsquared, Press etc. will be calculated in the transformed data. Most likely you will have to live with retransforming the data to its original measurement for purposes of management reporting. What does the invert of a payment or its fifth root mean to a VP of Finance?

0
#164041

High school student
Participant

“Sounds to me like you are not a Black Belt or trained in Lean or Six Sigma”. There are high school student who can run a regression. There’s no big deal to it.

0
Viewing 8 posts - 1 through 8 (of 8 total)

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