Experimental Design Vocabulary

Variable: Some condition of a system that can be altered or manipulated to learn how the system works. The specific variable being tested in an experiment is designated as the experimental variable.
Experimental Variable (Independent Variable): The single variable altered between treatment groups to test its specific impact.
- Example: Washing hands.
Dependent Variable: The variable measured to determine the results or outcomes of an experiment.
- Example: Rates of childbed fever.
Controlled Variable: Everything kept identical across all treatments to ensure fair comparison.
- Examples:
- Type of bed patients use.
- Amount of time patients are attended for.
- Age range of patients.
Treatment: A group or setup in which all subjects share the same specific conditions.
- Example: All patients assisted by physicians with washed hands constitute the "washed treatment," while all patients assisted by physicians with unwashed hands constitute the "unwashed treatment."
Controlled Treatment: A setup where conditions remain unchanged and the experimental variable is not altered (e.g., unwashed hands). An experimenter can repeat a treatment with known results or create a setup with a critical component missing so that no results occur (e.g., placebo treatments). Both variations qualify as controlled treatments.

When evaluating a hypothesis regarding how $A$ impacts $B$ :

Variables Component:
- Experimental Variable (Independent Variable): Represented by $A$ .
- Dependent Variable: Represented by $B$ .
- Controlled Variables: All other external variables maintained at constant levels.
Treatments Component:
- Experimental Treatment: Setup where condition $A$ is changed or altered.
- Controlled Treatment: Setup where condition $A$ remains unchanged.
- Negative Control Treatment: Design where $B$ should not be affected.
- Positive Control Treatment: Design where $B$ should be affected in a manner previously observed.

Test One Experimental Variable at a Time: Test only one experimental variable (independent variable) per trial to isolate its specific effect.
Multiple Treatment Groups: Include more than one treatment group in the experimental design.
Control Treatment: Always incorporate a control treatment for baseline comparison.
Comparative Analysis: Directly compare outcomes between control treatments and experimental treatments.
Mitigate Bias and Confounds: Systematically reduce potential sources of bias and confounding variables.
Raw Data Collection: Rigorously collect raw data throughout the process.

Definition and Core Principle: Control treatments and experimental treatments should differ strictly by the single experimental variable being tested.
- Example of a Confound: Placing all younger patients in one clinic and all older patients in another clinic introduces age as a confounding factor.
Strategies to Reduce Bias and Confounds:
- Sufficient Sample Size: Utilize more than one sample per treatment group. Comparing a single patient treated by a clean physician against a single patient treated by an unwashed physician is insufficient, as individual patients may possess unique susceptibility or resistance to infection.
- Repetition of Experiments: Repeat trials to verify consistency of observed outcomes.
- Randomized Sampling: Randomly assign subjects to treatment groups to eliminate systematic selection bias.
- Double-Blind or Masked Setups: Implement double-blind or masked protocols where participants and/or experimenters do not know treatment assignments, preventing unconscious bias.

Requirement for Data Collection: Every experimental design requires a precise, quantitative or qualitative method to measure the impact of the experimental variable on the dependent variable.
Semmelweis Case Study: In Ignaz Semmelweis's investigation of hospital infections, data collection involved recording whether or not individual patients contracted childbed fever following attendance by physicians under different hygiene conditions.