Detailed STAT 503 Lesson 1 - Introduction to Design of Experiments

A complete set of vocabulary flashcards covering Lesson 1 of STAT 503, including historical figures, experimental principles, and planning steps.

STAT 503 focus

The course focuses primarily on experimental design rather than statistical analysis and is described as more conceptual than math-oriented.

STAT 501 and STAT 502

The listed prerequisites for STAT 503, covering Regression Methods and Analysis of Variance respectively.

Montgomery, D. C. (2019)

The author and year of the primary course text: Design and Analysis of Experiments, 10th Edition.

Scientific method - first step

The stage where you decide what phenomenon you wish to investigate.

Scientific method - factor manipulation

The stage where you specify how the factor under study can be manipulated.

Scientific method - control extraneous conditions

The act of holding other conditions fixed so they do not influence the response being measured.

Scientific method - response measurement

Measuring the chosen response variable at several settings of the factor, including at least two settings.

Cause-and-effect conclusion

A relationship supported when changing a factor causes a phenomenon to change under controlled conditions.

Comparative experiment

An experiment designed to compare conditions, such as a treatment group and a control group.

Treatment-control factor structure

A structure where a treatment group and control group represent one factor with two levels.

One-factor-at-a-time strategy

An inefficient strategy where every factor is held constant while varying only one factor.

Engineering experiment purpose - time

To reduce the time required to design and develop new products and processes.

Engineering experiment purpose - performance

To improve the performance of existing processes.

Engineering experiment purpose - reliability

To improve the reliability and performance of products.

Engineering experiment purpose - robustness

To achieve products and processes that are robust.

Engineering experiment purpose - evaluation

To evaluate materials and design alternatives and set component/system tolerances.

Robustness in statistical analysis

A technique that is not overly influenced by bad data or outliers and still produces an appropriate answer.

Process robustness

The idea that a process should continue to work despite variation in who or what is involved.

General process model

An experimental process that consists of inputs, controllable factors, uncontrollable factors, and an output/response.

Inputs

Materials or quantities entering a process, such as ingredients in a recipe.

Controllable factors

Experimental factors whose settings can be controlled, such as baking time, temperature, or pan geometry.

Uncontrollable factors

Factors affecting the outcome that the experimenter cannot directly control.

Output / response

The measured result of the process or experiment.

R. A. Fisher

A major early figure (1918-1940s) whose work in agriculture established the foundations of modern experimental design.

Fisher-era developments

The introduction of factorial designs and analysis of variance (ANOVA).

Frank Yates

A colleague of R. A. Fisher who helped develop many concepts and procedures used in experimental design.

Orthogonal designs

Basic experimental ideas developed during the Fisher/Yates era of the 1920s to 1940s.

Latin squares

Concepts tracing back to early DOE work by Fisher and colleagues.

Sequential analysis

A methodology developed during World War II, initially used to improve the accuracy of long-range artillery guns.

First industrial era

A period from approximately 1951 through the late 1970s associated with response-surface methodology.

Box and Wilson

Key figures of the first industrial era associated with response-surface methodology.

Box and Wilson 1951 contribution

A key paper that treated output as a response function and sought optimum operating conditions.

George Box

An important statistician in response-surface methodology who married R. A. Fisher's daughter.

Second industrial era

A period from the late 1970s through 1990 associated with widespread quality-improvement initiatives.

CQI

Continuous Quality Improvement, a management goal that became prominent during the quality revolution.

TQM

Total Quality Management, a management technique associated with the statistical quality revolution.

W. Edwards Deming

A statistician who brought the importance of statistical quality control to Japan in the 1950s.

Taguchi

A Japanese engineer associated with orthogonal arrays, robust parameter design, and process robustness.

Orthogonal arrays

Experimental-design structures developed by Taguchi that are similar to Western fractional factorial designs.

Fractional factorial designs

Western terminology for designs very similar to Taguchi's orthogonal arrays.

Robust parameter design

A Taguchi-associated approach focused on choosing parameter settings that make processes or products robust.

Modern DOE era

The era beginning around 1990, driven by globalization and economic competitiveness.

Six Sigma

A quality-oriented approach popularized around 1990 that uses statistics, quality measures, and feedback loops.

Randomized double-blind clinical trial

The gold standard for medical product approval, essential for eliminating bias found in anecdotal evidence.

Randomization

The assignment of treatments to experimental units using a random process.

Experimental unit

The unit to which a treatment is assigned in an experiment.

Replication

Repeating observations or treatment applications across experimental units to estimate uncertainty and improve precision.

Standard error of the mean

The square root of the estimated variance of the sample mean.

Standard-error formula

$SE(\bar{x}) = \frac{s}{\sqrt{n}}$ or $\sqrt{\frac{s^2}{n}}$ under the usual independent-sample setting.

Effect of increasing $n$

As sample size increases, estimates of the mean become less variable and the standard error decreases.

Blocking

A design technique that incorporates factors responsible for undesirable variation so their contribution can be accounted for.

Nuisance factor

A factor that is not the primary scientific interest but contributes variability that must be addressed.

Blocking and noise

The process of explaining some experimental noise through blocks so unexplained error variance is smaller.

Treatment factor

A factor of primary scientific interest in an experiment.

Multi-factor design

An experimental design that studies combinations of multiple factors simultaneously.

Interactions

Relationships in which the effect of one factor depends on the level of another factor.

Confounding

A situation where the effects of two factors are mixed together so their separate contributions cannot be distinguished.

Intentional confounding

Deliberately mixing effects not of interest to gain efficiency for effects that matter, such as primary main effects.

Planning step 1

Recognition and statement of the problem.

Planning step 2

Choice of factors, levels, and ranges.

Planning step 3

Selection of the response variable(s).

Planning step 4

Choice of design.

Planning step 5

Conducting the experiment.

Planning step 6

Statistical analysis.

Planning step 7

Drawing conclusions and making recommendations.

Blocking factors

Nuisance factors explicitly incorporated into a design to prevent them from obscuring treatment effects.

Experimental factor

A factor whose levels can be specified/set by the experimenter and randomly assigned to experimental units.

Classification factor

An inherent characteristic/attribute of an experimental unit that cannot be changed or randomly assigned.

Quantitative factor

A factor for which specified numerical levels can be assigned, such as pH level or concentration.

Qualitative factor

A categorical factor consisting of different types, such as plant species, brand, or gender.

Sequential experimentation

An iterative process where current knowledge informs the design of the next experiment.

Learning cycle of experimentation

The process where findings from one experiment become part of the knowledge base for subsequent designs stage.

Randomization memory rule

Random assignment protects against systematic treatment-allocation bias.

Replication memory rule

More independent replication improves estimation of uncertainty and usually reduces standard error.

Blocking memory rule

Explain nuisance variation through blocks so less variation remains in experimental error.