Receive a call from process engineering manager, she request me to arrange statistical analysis training for her process engineers. She wants the engineers to have better understanding of statistical analysis and apply it in their daily work.
I am willing to customize training for my colleague, but I don’t know what to deliver. I think of this problem for the past few days, nothing finalized.
Tonight, I have a chance to really sit down to review my old training material and summarize something which is useful for them.
Step 1 : Understand your process
All process deliver result, result is the outcome of combination of various process input variable. The relationship between input and output could be translate into mathematical term as below:
If you do not understand well the relationship of input and output variable, you may end up collecting data which is not correlated.
Step 2 : Establish problem statement
Click here to link to previous post talk about problem statement establishment
Step 3 : Objective setting
Objective only can be set full understanding of problem statement.
If you don’t know your problem well, you don’t what to focus, and you may have problem to set a right objective.
Once a clear objective set, you will know the information should be collect for statistical analysis.
Step 4 : Establish data collection plan
Plan your data collection schedule or interval. In most cases, data collection for manufacturing process should be carry out in a stable process condition.
Anyway, there are cases where process condition was purposely tweak into unstable condition for data collection purpose.
It it depend on the objective of the data collection.
Once you have the data collection. Here come to the part of statistical analysis
Step 5 : Summarize your data
Most statistical analysis software (such as Minitab) provide feature of generate data summary report.
Information can be obtain from the summary report are
Mean / median
Data distribution and spread
Normality test
Confident interval
Standard deviation
And others
The report provide a rough idea of your data behavior.
Step 6 : Check for equal variance
Check for equal variance is recommended if data was collected from various sub group, the purpose of the assessment is to understand the difference between each data set.
Box plot is the common graphical method I like to use for quick assessment.
I may apply Barttlett’s test or Levene’s test if I would like to have proper statistical assessment.
Step 7 : Process stability check
Main purpose of process stability assessment is to understand the process condition when data was being collected.
Time serious, I-MR and X – Bar R chart are the most commonly use graphical presentation in most process engineering.
The assessments provide an idea of process behavior over time.
Step 8 : Process capability analysis
Process capability analysis tells a process capability to produce output that meeting predefined target.
Click here to link to previous post talk about process capability analysis for normal data
Click here to link to previous post talk about process capability analysis for non normal data
Conclude your finding based on information obtain from Step 5 – 8
