Thursday, April 12, 2012

Box-Cox Transformation

Minitab procedure
Stat  >  Control Chart > Box-Cox Transformation

The Box-Cox transformation is used to transform non-normal data or when within subgroup variation is unstable that cause the data non-normal. The Box-Cox procedure was developed to determine an optimal transformation from the collection of the form

The optimal value of Lambda is then used as a guide for data transformation. It is not necessary to use precisely define Lambda. For example, use Lambda = 0.5 instead of Lambda = 0.48339.

Common value for Lambda include

You may use the comman to find the optimal Lambda value and choose to store the transformed data with the command, do not select the Box-Cox option when you make a Control Chart, do ing so will cause double transform of the data.

Example of Box-Cox Plot

Ref : Minitab Stat Guide

Friday, April 6, 2012

Run Test

I learn something new from the past Kulim training.

A Green Belt shared with me how we can apply Run Test to test if the sample series is distributed randomly.
Test procedure as below

Minitab >> Stat >> Nonparametrics >> Run Test

Note : run test is a nonparametris test because we make no assumptions about the distribution of the population being sampled

Session Window

Runs test for C1

Runs above and below K = 10.0000

The observed number of runs = 17
The expected number of runs = 15.7333
17 observations above K, 13 below
P-value = 0.632

P-value > 0.05, the data being test is no significant different evident to proof the data is not in random order.