Experiment Study Guide

📖 Core Concepts Experiment – A purposeful procedure that manipulates at least one factor to test a hypothesis or evaluate something new. Control – A condition that is kept constant (or a known outcome) to isolate the effect of the independent variable. Independent Variable (IV) – The factor the researcher deliberately changes. Dependent Variable (DV) – The outcome that is measured to see how it responds to the IV. Null Hypothesis (H₀) – The default claim that there is no effect (no difference between groups). Random Assignment – Placing subjects into treatment or control groups by chance to neutralize confounding factors. Double‑Blind Design – Neither participants nor experimenters know who receives the treatment, preventing bias. --- 📌 Must Remember Experiments cannot prove a hypothesis absolutely; they can only add support or provide disproof. A single counterexample can falsify a theory, though the theory may be revised. Positive control verifies the system can produce a response; negative control establishes the baseline. Randomization creates equivalent groups on average, minimizing bias. Average Treatment Effect (ATE) = mean outcome (treatment) – mean outcome (control). Ethical rule: Human experiments require informed consent and must not expose participants to harmful or substandard treatments without justification. --- 🔄 Key Processes Designing a Controlled Experiment Identify IV and DV. Set up positive and negative controls. Decide on replication (duplicate/triplicate). Randomly assign subjects to treatment vs. control. (If human) Apply double‑blind procedure. Testing a Hypothesis State H₁ (research hypothesis) and H₀ (null). Run experiment, collect DV data. Compare results to predictions of H₀ (statistical significance). Conclude support for H₁ or retain H₀. Conducting a Meta‑Analysis Gather separate experimental studies on the same question. Extract effect sizes and variances. Combine statistically to produce a pooled estimate with tighter confidence intervals. --- 🔍 Key Comparisons Experiment vs. Observation Experiment: deliberate manipulation of at least one factor. Observation: passive data collection; no manipulation. Controlled Lab Experiment vs. Field Experiment Lab: artificial setting, high control over variables. Field: natural setting, lower control, higher ecological validity. Positive Control vs. Negative Control Positive: known to give a positive result → confirms system works. Negative: known to give a negative result → defines background/noise level. True Experiment (Random Allocation) vs. Quasi‑Experiment True: subjects randomly assigned → eliminates selection bias. Quasi: allocation not random → higher risk of confounding. --- ⚠️ Common Misunderstandings “An experiment proves a theory.” – It only provides support; proof is never absolute. “If results are not significant, the hypothesis is false.” – Non‑significant results mean we cannot reject H₀; they don’t prove H₀ true. “Controls are optional.” – Without controls you cannot attribute observed changes to the IV. “Observational studies are just “less rigorous” experiments.” – They lack manipulation and are prone to selection bias, making causal inference weaker. --- 🧠 Mental Models / Intuition “Cause‑and‑Effect Filter” – Think of the experiment as a sieve that lets only the IV’s effect pass through, while controls block every other influence. “Randomization as a Mixer” – Shuffling subjects before assigning groups ensures any hidden traits are evenly spread, like mixing ingredients before baking. “Null hypothesis as a baseline thermostat” – H₀ sets the temperature at “no change”; any deviation that’s big enough triggers the alarm (statistical significance). --- 🚩 Exceptions & Edge Cases Ethical constraints may force researchers to use observational or field studies even when randomization is ideal. Small sample sizes can produce random imbalances despite randomization; replication becomes critical. Double‑blind infeasibility (e.g., surgery vs. medication) – alternative blinding or objective outcome measures are needed. --- 📍 When to Use Which Use a controlled lab experiment when you need tight variable control and can manipulate the IV safely. Use a field experiment when ecological validity matters and you can still randomize subjects. Use an observational study when manipulation is unethical or impractical (e.g., smoking exposure). Apply double‑blind design for human studies with subjective outcomes (pain scores, self‑report). Choose meta‑analysis when multiple small studies exist and you need a stronger overall estimate. --- 👀 Patterns to Recognize Presence of a control group → likely a controlled experiment. Random assignment mentioned → indicates a true experiment. “Informed consent” or “IRB” → human subject research, check for double‑blind or ethical safeguards. “Average treatment effect” language → statistical comparison of treatment vs. control groups. “Field” or “natural setting” → look for potential confounders and reduced control. --- 🗂️ Exam Traps Distractor: “Experiments can prove hypotheses.” – Wrong; they only provide support. Distractor: “A non‑significant p‑value proves the null hypothesis.” – Incorrect; it merely fails to reject H₀. Distractor: “Observational studies are just less expensive experiments.” – Misleading; they lack manipulation and have different bias profiles. Distractor: “Negative controls are unnecessary if you have a positive control.” – False; both establish baseline and system functionality. Distractor: “Randomization eliminates all bias.” – Overstated; it addresses selection bias but not measurement or reporting bias.