 |
 |
Testimonials
Do you need to use statistics to become a good Black Belt?
If I'm afraid of learning and using statistics, can I carry out improvement projects without using them?
Most people have come across statistics at some time in their academic past and for many it has come across as a difficult subject of little practical use. It is not surprising, therefore, that many people would prefer not to have anything to do with it as they move into the intensely practical world of Business.
This can also put people off Lean Six Sigma training and leads them to wonder whether they can get by "without the statistics" as a Black Belt, and to question the need to use statistical tools in problem solving.
So, do we need to use statistical tools to solve problems, or can a Black Belt really get by without using statistics?
To answer this question properly we need to understand what we mean by 'statistics'. Look up the word 'statistics' in the dictionary and you will find something along the lines of "science that deals with the collection, classification, analysis, and interpretation of numerical facts or data..."
It is a large subject, and it's pretty unlikely that you are going to solve a problem of any substance without having to resort to the collection, analysis and interpretation of numerical facts (statistics). As soon as you have worked out an average, for example, you are in fact using a statistical tool. Most people are comfortable with using averages, and discussing maximum and minimum values
So if statistics is a broad subject and at least some of it is 'OK' for most of us, then let's break the subject down into its elements to find out what exactly we are afraid of, and find out if it is something that must be used to be a successful Black Belt.
Statistical tools can be broken down into three broad areas or branches. These are:
- Descriptive (or basic) statistics
- Inferential statistics
- Analytical statistics
Descriptive statistics consist of ways to summarise, or describe, your data. These are very commonly used and it is difficult to image life in the modern world without them. Just listen to the news and you will be bombarded with them: averages, medians, maximums, minimums, ranges, and even sometimes standard deviations.
Maximums and minimums are used to describe the extremities of the data, averages, or more precisely 'means', and 'medians' are used to describe the centre of the data (where the data is located). Range and standard deviations describe the variability in the data.
When we describe one solution or condition as better than another we usually mean 'on average' whether we actually state this or not. We cannot really avoid using descriptive statistics in business and as they are usually taught quite successfully at high school, and we often use them without realising it, no one should be afraid of this branch of statistics.
Inferential statistics are sometimes a little more complex and are more likely to be a source of anguish. We need to move into this branch of statistics when we are unable to collect all the data in a population and need to resort to 'sampling'.
Say, for example, you were interested in comparing teacher morale in State verses Private schools. There is no way that you could interview or survey all of the teachers in each sector and so you would be forced to take samples.
You ask a sample of teachers certain questions related to morale, work out average responses for each sector and compare them to see if there are any differences. Now if the differences in sector averages are large this allows you to safely draw conclusions from your sample data, but as these differences reduce there comes a point when you wonder if the differences you are noting between sectors are in fact significant. No two schools are the same, and even different teachers in the same school are going to have differences in opinion. The differences you are seeing could just be due to the individual teachers you have chosen in your sample!
This is where hypothesis testing comes into play. Hypothesis testing tools allow you to calculate the likelihood, or probability, that any detected difference could come from error in the sampling. If the probability of error is low then it is safe to assume that the results you are seeing are real and do represent real differences. If the probability of error is high, then we take the conservative perspective and assume that 'no significant difference' exists.
Take a moment to reflect on how often you will need to rely on samples when taking decisions, either because the populations are too large or because you only have data over a relatively short time period - this is nearly all the time!
Thus whenever you are forced to rely on samples and you are looking to compare two or more things you will need to rely on this branch of statistics to understand the risk of making incorrect decisions. If the differences are very large then it may be obvious that the risk is low, and so there is no need to make use of hypothesis testing. This is likely to be true when you are tackling relatively easy problems where 'the fruit is on the floor' as the analogy goes.
You can get away without inferential statistics, but only if you are looking to find obvious root causes and make improvements where the solution is relatively clear. Will this always be true for a Black Belt? I think not, for anything other than basic problem solving a good understanding of statistics is required.
Analytical statistics refer to model building and simulation tools such as Multiple Regression and Monte Carlo simulation. Multiple Regression is the tool Black Belts use to build the transfer function, or equation that explains the relationships between inputs and outputs in a process. Monte Carlo simulation is a tool used to help with tolerancing and sensitivity analysis. Now it is probably fair to say that in many projects, especially transactional ones, you won't need to use these types of tools. They may be useful as your projects become more technical.
So what can we conclude from this examination of statistical tools, and can you get away without them? It is clear that there are many descriptive tools which are rather basic and which you cannot do without, but these are less likely to cause you difficulties.
So what about inferential statistical tools? Any self respecting Green or Black Belt should have these tools in their tool box - to be used where necessary. As a Green or Black Belt you should be able to tackle more difficult problems where the root causes and solutions are not clear, and an understanding of risk is important.
The analytical toolset is more complex, and you may well get away without them, especially if you are doing transactional projects where root causes are easy to find, or where subjective tools such are process mapping, fishbone diagrams and FMEA will do.
At SigmaPro we teach Yellow Belts and foundation level students how to carry out relatively straight forward projects without inferential statistical tools, but Green and Black Belts are taught inferential statistics. Once students have an understanding of the tools, then it is surprising how many projects use them.
Author
Mike Titchen - Six Sigma Master Black Belt (view more about Mike Titchen)
Quick Enquiry or Call Back
Get in touch, ask us anything! How can we help?
Six Sigma Yellow Belt | Six Sigma Green Belt | Six Sigma Black Belt | Six Sigma Master Black Belt
Locations : Lean Six Sigma Training in Birmingham | London | Newcastle | Leeds