Experiment Objective, Factor, Level and Response

Why it matters

When both the clamping arrangement and the processing time are changed on the same day in a process, you cannot tell from the result table alone which change produced the better product. The experiment objective poses the question that reduces this uncertainty. Instead of a broad phrase such as better production, it asks how two selected factors change the bond strength of a defined product under a defined acceptance condition. Once the objective is written, the decision must also be visible: which condition is worth trying, which result is insufficient, and in which situation will the work be stopped? Posing the question clearly comes before collecting more numbers. The property to be measured gets a definition that is independent of the expectation of the person planning the experiment. Past records can show which factors may be investigated; but if the conditions changed together in the past, these records alone are not evidence of cause. A controlled experiment places the selected changes into comparable runs. For this reason a suggestion drawn from past data and a result tested by experiment are not reported under the same name. The purpose of the experiment is also not to explain every possible cause at once; it is to produce the information that will decide a clearly bounded question.

A factor is a condition that is deliberately changed during the experiment. A level is a value of that factor that is tried. The response is the result measured after the condition is applied. Temperature may be the factor, a selected temperature value the level, and bond strength the response. Do not use the same concept with different meanings in different columns. If the ambient temperature is recorded but not changed, it is not a factor studied in this experiment; it is a record of a condition that may affect the result. Being changeable is also not enough by itself. The factor must be applicable safely, its real value must be verifiable, and it must be possible to maintain it in later production. Whether the setting on paper and the condition actually applied are the same is checked separately.

Choosing the primary response before the trial limits the tendency to look for a successful-looking result afterwards. When no difference appears in strength, moving to cycle time, and when none appears there, moving to energy and declaring the work a success, is not testing the same question. Other properties can be monitored; but their roles are assigned in advance. The primary response answers the main question of the experiment, safety and quality criteria limit acceptance, and records kept for exploration may give rise to new questions. This separation does not devalue the secondary information. It keeps clear which decision rests on which evidence. It also makes visible when the original objective is changed because the result was not liked.

The suitability of the measurement system is a precondition of the experiment. If the measurement variability is larger than the expected effect of the setting change, small differences become hard to interpret. Rereading the same specimen and applying the same condition again to a new specimen produce different information. The resolution, bias and stability checks learned in the previous module ask here whether you can really see the difference you are looking for. An experiment does not solve a measurement problem by itself. In addition, safe levels are not chosen by calculation alone; the technical limits of the work and the protective conditions are kept. A high response obtained at an unsafe point is not an applicable option.

Formula