What is Design of Experiment?

Why Do Designs of Experiments Fail?

The primary reason is the failure to follow a systematic procedure to design the experiment. A systematic procedure for designing an experiment is provided in Table 1. The four-steps process of designing an experiment follows a very natural and logical order that has been practiced by humans, animals, and any living entities for that matter. Therefore, these four-steps are common to any research, regardless of disciplines. These steps have been adopted in research articles, journals, theses, dissertations, etc. In other words, the most common design of experiments, regardless of discipline, has been adopted for the statistical design of experiments too. This simple intuitive natural process includes four major areas such as (1) making a hypothesis, (2) choose an appropriate method to prove the hypothesis, (3) produce results by analyzing the data collected following the method, and finally (4) drawing conclusion in the context of the problem.

Activities performed in each step

  • Formulation of the question(s) by performing thorough existing literature investigations
  • The question of interests must be legitimate and validated by existing published research
  • Translating the formulation of the research questions into statistical hypotheses
  • Determine both dependent and independent variables those are related to the hypothesis
  • Select an appropriate statistical method based on the related variables with the hypothesis
  • Detail descriptions of the experimental procedure
  • Randomization, Replication, and Blocking
  • Define the experimental and observational units
  • Define the data collection procedure
  • Data analysis plan (software use, etc.)
  • Produce results by statistical analyses of the data collected following the detail procedure specified in step 2 method section
  • Post-hoc analysis when appropriate
  • Descriptive statistics results
  • Inferential statistics results
  • Statistical explanation of the results
  • Data diagnostic Analysis to check the data quality
  • Interpret the results in the context of the problem, meaning that translating the statistical findings into the formulated research question in step 1 for general audience.
  • Formulation of new hypotheses (future studies). Whether the hypothesis is proved or disproved, the new information, knowledge gained would produce some new directions and formulation of new research questions.
  • Going back to step 1 with well informed about the situation to produce a better formulation of the research questions and continues until satisfaction (never happen!). We wouldn’t have to write this book again while a ton great books on DOE exit today.
  • Therefore, this is a refining process
  • Development of new theory