Design of experiments (English: design of experiments, DOE).

The quality of the product is mainly determined by the design. A good experimental design includes several aspects.

The first is to clearly measure the indicators of product quality. 6σ management emphasizes using data to speak, so this quality indicator must be an indicator that can be quantified. In the experimental design, it is called the experimental indicator, also known as the response variable ( response variable) or output variable.

The second is to find possible factors that affect the test indicators, also known as influencing factors and input variables. The various states of factor changes are called levels, and it is required to initially determine the range of factor levels based on professional knowledge.

The third is to choose an appropriate experimental design method based on actual problems. There are many methods of experimental design, and each method has different applicable conditions. If you choose a suitable method, you can get twice the result with half the effort. If you choose an incorrect method or do not conduct an effective experimental design at all, you will get twice the result with half the effort. .

The fourth is to scientifically analyze the test results, including intuitive analysis of data, variance analysis, regression analysis and other statistical analysis methods. These tasks can be completed with the help of Minititab software.

Major issues in experimental design include determining validity, reliability, and replicability. For example, these issues can be partially addressed by carefully selecting independent variables, reducing the risk of measurement error, and ensuring that the documentation of methods is sufficiently detailed. Relevant issues include achieving appropriate levels of statistical power and sensitivity.

Extended information:

Statistical control

It is best to carry out reasonable statistical control of the process before conducting the designed experiment. If this is not possible, the experiment can be carefully designed with appropriate blocking, replication, and randomization.

To control for nuisance variables, researchers developed control checks as additional measures. Researchers should ensure that uncontrolled effects (e.g., perceived source credibility) do not distort study results. A manipulation check is an example of a control check. Operational checks allow investigators to isolate key variables to strengthen support that these variables are operating as planned.

One of the most important requirements of experimental research design is that the influence of spurious, intermediate and a priori variables must be eliminated. In the most basic model, a cause (X) causes an effect (Y). But there might be a third variable (Z) that affects (Y), and X might not be the real cause at all. It is said that Z is a dummy variable and must be controlled.

The same is true for intermediate variables (variables between the presumed cause (X) and the effect (Y)) and a priori variables (variables preceding the presumed cause (X) to be the real cause). When a third variable is involved and has not been controlled for, the relationship is called a zero-order relationship. In most practical applications of experimental research designs, there are multiple reasons (X1, X2, X3). In most designs, only one of these causes can be addressed at a time.

Reference: Baidu Encyclopedia-Experimental Design