Scientific Method

Definition: the variable that the experimenter deliberately changes to observe its effect.
In the transcript example: the independent variable is the fertilizer.
Example framing from the talk: two groups
- Group 1: with fertilizer
- Group 2: without fertilizer (the “no fertilizer”/control group)
Mathematical framing (conceptual):
- Let X \in {\text{fertilizer}, \text{no fertilizer}} denote the treatment assignment.
- The effect is studied by comparing the outcome across levels of X.

Definition: the variable measured to assess the effect of the independent variable.
In the apples example: the dependent variable is the amount (yield) of apples produced.
In the cancer example from the transcript: the dependent variable is the tumor size.
Conceptual formula: the study observes how the outcome Y changes as a function of X.
Measurement notes from the transcript:
- For apples: weighing the apples (yield) is the measure; consider whether to count apples or weigh total yield.
- For tumors: size is the measure; the choice of metric matters (size units, imaging, etc.).

Definition: variables not controlled in the study that can influence the dependent variable and confound results.
Examples mentioned:
- Weather conditions (rain, temperature) affecting apple yield or plant health.
- Natural factors like pests, disease, or fruit rot affecting measurements.
- The transcript questions: “which ones do you count?” (selection bias in what is counted).
- In the tumor study: reaction to treatment, side effects, and other patient responses can introduce variability if not controlled.
Role in interpretation: uncontrolled variables reduce internal validity and can obscure the effect of the independent variable.

Definition: variables kept constant across all experimental groups to isolate the effect of the independent variable.
Examples from the transcript:
- Amount of water given to the plants.
- Soil conditions or dirt/soil type (mentioned as “soil” or “dirt”).
- Starting conditions (baseline health, initial size, etc.).
- In the cancer example: the amount/dose of chemotherapy kept the same for all groups.
- Age range or other participant characteristics discussed as factors to control (age is mentioned as something to consider controlling).
- Adherence/compliance: “they actually stick with the medicine” is a factor to control or at least monitor (to reduce variation).
- Wearables or monitoring devices (age, activity) suggested as potential controls.
Purpose: by holding these constants, any observed differences in the dependent variable can be more confidently attributed to the independent variable.

Measurement choices impact results and interpretation.
- For apples: weighing vs counting; deciding whether to include rotten or damaged fruit.
- For tumors: ensuring consistent measurement technique and timing across groups.
Data quality topics reflected in the transcript:
- Consistency in counting or measurement method.
- Whether measurement bias is present (which fruits are counted, how tumors are measured).
Possible simple quantitative framing:
- If there are two groups, ranges of outcomes can be summarized by means
  \bar{Y}{\text{fert}} \quad\text{and}\quad \bar{Y}{\text{no}}
- Difference in means: \Delta = \bar{Y}{\text{fert}} - \bar{Y}{\text{no}}
- If performing a basic comparison, a test statistic (conceptually) might be
  t = \frac{\Delta}{\text{SE}(\Delta)}
  where SE(Δ) is the standard error of the difference in means.

Objective: Assess whether adding fertilizer affects apple yield.
Independent variable: X \in {\text{fertilizer}, \text{no fertilizer}}.
Dependent variable: Apple yield, measured as weight or count of apples produced.
Experimental groups: at least two groups corresponding to the two levels of X.
Controlled variables (to improve internal validity):
- Water amount, soil type/quality, sunlight exposure, and other growing conditions.
- Start conditions: initial tree size, health, and baseline yield potential.
- Measurement method: consistent weighing or counting method; decision on whether to include damaged/rotten fruit.
Uncontrolled variables to be monitored or noted: weather fluctuations, pests, disease, and other natural variability.
Data considerations: ensure random assignment to groups if possible; decide on sampling method for apples to measure yield.

Objective: Compare tumor size under different treatment conditions with a fixed chemo regimen.
Independent variable: treatment type or strategy (e.g., different agents or regimens). In the transcript, the discussion focuses on keeping the chemo dose constant across groups, implying the treatment variable is other than dose (or a fixed-dose approach).
Dependent variable: tumor size (how the tumor grows or shrinks over time).
Controlled variables (critical in clinical studies):
- Same amount/dose of chemotherapy across all groups.
- Baseline characteristics: age range, initial tumor size, overall health status.
- Compliance: ensuring patients adhere to the treatment plan (“they actually stick with the medicine”).
- Monitoring tools: wearable data mentioned as a potential way to account for activity or health status.
Uncontrolled variables to be anticipated: individual differences in reaction to treatment, side effects, metabolic differences, and other biological factors.
Additional notes:
- The transcript touches on practical considerations like starting conditions and ongoing adherence as factors that can influence outcomes.
- When analyzing results, one would consider controlling or adjusting for these factors to isolate the effect of the independent variable on tumor size.

Possible roles mentioned in the transcript:
- Engineer
- Data analyst
Practical interpretation: in real-world projects, teams decide who handles design, data collection, and analysis; clear definitions of variables and consistent data collection are foundational to any role.
Real-world relevance: understanding experimental design helps in agriculture, medicine, and any field where evidence of causation is sought.

Core idea: well-designed experiments aim to establish causal relationships by manipulating one independent variable while controlling others.
Key commitments:
- Randomization (to balance unknown confounders, if possible).
- Control of extraneous variables to isolate the effect of the independent variable.
- Clear, measurable dependent variable and transparent measurement methods.
- Consideration of ethical, practical, and logistical factors in real-world contexts (e.g., clinical settings).

The transcript does not explicitly discuss ethics, but clinical contexts raise common considerations: patient safety, consent, and approvals when studying treatments.
Practical implications include the importance of adherence monitoring, potential biases from measurement choices, and the need for transparent reporting of methods and variables.

Independent Variable (IV): the factor deliberately changed. Example: fertilizer vs no fertilizer. X \in {\text{fertilizer}, \text{no fertilizer}}
Dependent Variable (DV): the outcome measured. Example: apple yield; tumor size. Y = \text{yield} \quad\text{or}\quad Y = \text{tumor size}
Controlled Variables (CV): conditions kept constant across groups (water, soil, dosage, age range, etc.).
Uncontrolled/Extraneous Variables (UV/EV): weather, pests, measurement bias, adherence variability, etc.
Measurement and data quality: choose clear, consistent measurement methods to minimize bias.