DSC6008 Chap.1 Experimental Questions and Design Vocabulary
Experimental Questions and Design Vocabulary
Build the experiment in words first
A practical aim becomes testable only after it names a response, the factors to vary, their levels, the resulting treatments and the units that receive those treatments independently. This vocabulary is structural rather than decorative. It determines the run count, the meaning of replication and the population to which a conclusion can apply.
Nuisance variables are separated from focal factors so they can be recorded, randomized over or controlled through blocking.
Move from an objective to a measurable response
An objective should name one practical comparison and the decision it will inform. The response then needs a measurement unit, instrument or scoring rule.
A phrase such as better quality is not yet a response because two investigators could measure it differently. Once the outcome is operational, the factors can be chosen for their ability to answer the objective rather than because they happen to be easy to vary.
Build complete treatments from factor levels
A level is one specified setting of a factor, while a treatment is the full combination applied in a run.
With temperature at two settings and pressure at three settings, the crossed treatment set contains six combinations before replication. Stating only temperature high is incomplete when pressure also varies.
A treatment table makes missing combinations visible and distinguishes the number of unique conditions from the number of independently executed runs.
Protect the unit of analysis
Several measurements on one treated object do not automatically create several experimental units. The smallest independently assigned object controls the effective sample size.
Writing that unit beside the treatment combination prevents pseudoreplication before it reaches a spreadsheet or model formula. Subsamples can reduce measurement noise, but their dependence must remain visible in the analysis.
Separate nuisance conditions from focal inputs
Day, batch, operator or location may change the response without answering the research question.
Record those conditions before outcomes are known. Randomization can spread unknown influences across treatments, while blocking can keep comparisons local within a known nuisance group.
Turning every nuisance variable into a treatment factor is rarely practical, so the design statement should explain which inputs answer the question and which protect the comparison.
Finish with a reproducible assignment sentence
A complete description says which units are eligible, which treatment combinations exist, how assignments are chosen and what response is collected from each unit.
Another analyst should be able to construct the run table from that sentence without guessing. This final check catches vague factor levels, hidden repeated measurements and treatment labels that omit part of the condition.
What this chapter covers
- 01
Research objectives
- 02
Response variables
- 03
Factors and levels
- 04
Treatment combinations
- 05
Experimental units
- 06
Nuisance variation
Worked example · free
Define a coating experiment before counting runs
- 1Name curing time as the measured response.
- 3Identify temperature and thickness as factors and each pairing as a treatment.
- 4Name the panel as the experimental unit and humidity as a nuisance variable.
Key terms
- Experimental Unit
- The smallest unit to which a treatment is independently applied.
- Treatment Combination
- One complete combination of the selected factor levels.
- Nuisance Variable
- A variable that can affect the response without being the target of the comparison.
Experimental Questions and Design Vocabulary FAQ
Why is the experimental unit so important?
It decides what counts as independent information. Several readings from one assigned object may refine that object's measured response, but they do not create additional treatment assignments. Using the reading count as the sample size would make the reported uncertainty too small and the conclusion too confident.
Is a factor the same as a treatment?
No. A factor is an input variable, and a level is one setting of that factor. A treatment is the complete combination of levels across every factor in the experiment. Two factors with two levels each therefore define four treatments before replication is added.
What should an operational response include?
It should name the quantity, its unit or scale, and the measurement procedure closely enough that another person could collect the same outcome. Words such as quality or improvement are not operational until the scoring rule, instrument or calculation is stated.
Assessment move
Rewrite every design question as a one-line assignment statement. Circle the response, underline the complete treatment and box the independently assigned unit. Then count treatments separately from runs. This small routine exposes ambiguous units and incomplete factor combinations before any formula is chosen.
Working through Experimental Questions and Design Vocabulary in DSC6008? Sia is AskSia’s AI Statistics tutor — ask any DSC6008 Experimental Questions and Design Vocabulary question and get a clear, step-by-step explanation grounded in how DSC6008 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
