If nothing was uncertain we would not need statistics.
Since nearly everything varies in some fashion, we need a way to describe and work with that variability.
We already know this and we know about statistics as being the right set of tools. Yet we hesitate, avoid, and refuse to pick up the appropriate tool.
Maybe I learned statistics in a good way.
Working in a factory as a manufacturing engineer with a boss that expected us to conduct an experiment every day.
We learned and practiced using data collection, measurement error calculations, hypothesis testing, statistical process control and design of experiments as part of our work.
We could see the advantages of these various tools.
We also were not allowed to present data without the appropriate statistical analysis. A proposal had to include how the data was to be collected and analyzed.
We spoke with data and included the appropriate statistical analysis. It was a part of the way we worked.
In hindsight I was lucky. Too many do too little with data.
The value of statistics
We all need to make decisions. Every day we consider the odds.
What is the chance of a better parking space closer to the door? What is the likelihood this failure will occur in other products? How much better is the product when using vendor A’s components?
We already use engineering judgment and experimental data to make decisions.
We can enhance our decision-making ability. We can reduce the uncertainty of our estimates.
And, we can consider our risks, all using statistics.
Yet, if we accept recommendations without the supporting statistics, then we allow ourselves to make poor decisions.
Encourage the use of statistics
Be like my former boss. Encourage and support the use of statistics.
This doesn’t mean you have to be the expert statistician, yet you can learn along with your team the tools you need to understand your data and improve your decision making.
For test proposals, you should expect:
Identification of data collection requirements and associated measurement errors
- Estimate of sample size required, including assumptions
- Type I and Type II error estimates
- Details on data collection method
- Details on data analysis methods
For test results, you should expect:
- The data and notes on the collection
- Decisions about anomalies (outliers) justified
- Enough detail that the analysis could be replicated
- Any additional assumptions and associated risk
- Statistical analysis approach and results
- Interpretation of the results
- Next steps
Basically expect the use of statistics to design, conduct and analyze experiments.
Do experiments to learn and reduce uncertainty. As my boss would often say, “let the data talk.”
You can do the same and enjoy the benefits of improving decision making across your organization.
Related:
Role of Reliability Statistics (article)
Why do Statistical Based Testing? (article)
Leave a Reply