Design of experiments is a strategic and tactical approach to experimentation. With much of six sigma, we are waiting for assignable-cause variation to exhibit itself. With design of experiments, we are manipulating various factors (chosen by the cross-functional team) and manipulating them at different levels to see their effect on some desired result.
A defect location check sheet (also known as a defect map or a measles chart) is a structured, prepared form for collecting and analyzing data that provides a visual image of the item being evaluated so that data can be collected visually rather than with words. A common practice in quality assurance or even in six sigma is to count the number of defects found. One could keep track of the defect rate, and maybe even use an attribute chart to monitor, control, and continuously improve upon defect rates.
Continuous data can be measured on a continuum. Think of it as being able to divide a measure by one half, and in half again, and in half again, – to infinity. Contrast continuous data with discrete data where there are only a finite number of values possible or if there is a space on the number line between two possible values. For example, it’s impossible to roll the 2.3 on a roll of the dice. You can either roll a two or three but nothing in between.
First, the team defines a problem in the define phase – the project charter, customer needs and requirements, and they developed a process map. The next phase is the measurement phase where the team develops a data collection plan and establishes a baseline sigma. Following the measurement phase is the analyze phase where the team performs data analysis, process analysis, and root-cause analysis. Next is the improvement phase where the team generates a host of possible solutions. They select a solution and then ultimately implement the solution. Finally, in order to maintain the solution, the team develops a control plan and a response plan as a part of the control phase of DMAIC.
The effort/impact matrix (also known as the impact/effort matrix) is used to help the team to decide which of the numerous solutions to implement appear to be the easiest (least effort) while having the most favorable impact.
The F-distribution is another distribution that is used extensively in statistical analysis to test certain hypotheses arising from the comparison of two or more normal distributions. One of the ‘operational definitions’ of quality follows: Quality is considered the minimum deviation around a target value. We often spend a significant amount of time and money making sure our processes are ‘targeted’ appropriately, but it is just as important to assure that processes are operating as consistently (that is, with as little variation) as possible.
A factorial experiment is the simultaneous manipulation of factors at different levels to see their effect on some desired result. A factorial experiment could be ‘a full factorial’ or it could be a ‘fractional factorial’ experiment. If the experiment is fractionated, the practitioner gives up something in the sense that main effects are aliased with interactions, or interactions are aliased with other interactions, and so forth. A full fractional factorial experiment has none of the aliasing, but it cost more because you have to run more experimental trials.
A flowchart (also known as a process flow diagram) is a graphical tool that depicts distinct steps of a process in sequential order (from top to bottom of the page). The basic idea is to include all of the steps of critical importance to the process. Also, flowcharts are often annotated with performance information.
Kanban is a Japanese term that literally means ‘sign-board.’ Kanban is a scheduling system for lean and Just-In-Time (JIT) production. According to its creator, Taiichi Ohno, kanban (a signal to produce) is a means through which JIT is achieved.
The alternate hypothesis is the complement of the null hypothesis. The null hypothesis is what you anticipate through randomness. The alternative hypothesis, sometimes known as the alternate hypothesis is the opposite of that. It is what you would not anticipate by randomness. More often than not you are trying to reject the null because you are trying to see a change in something – considering that most six sigma projects are trying to fix broken processes.
