Random Assignment: The process of allocating participants to experimental groups using chance, ensuring each participant has an equal probability of being assigned to any group.
- Number all 60 participants from 1 to 60
- Use a random number generator to create a sequence
- Assign first 30 numbers to treatment group, remaining 30 to control
- Document the assignment sequence for reproducibility
Participants: 1, 2, 3, ..., 60
Random number generator produces: 45, 12, 33, 7, 29, ...
First 30 in sequence → Treatment group, last 30 → Control group
Record assignment method for transparency and validation
Number participants 1-60, use random generator to assign first 30 to treatment and last 30 to control. This ensures equal probability and balances confounding variables.
• Equal Probability: Every participant has same chance of group assignment
• Chance Determination: Assignment based solely on random mechanism
• Balance Confounding: Randomization averages out differences between groups
Confounding Variable: A variable that affects both the independent variable (treatment) and the dependent variable (outcome), potentially creating spurious associations.
Prior knowledge affects both teaching method effectiveness and learning outcomes
Random assignment distributes prior knowledge equally across groups on average
With balanced prior knowledge, differences in outcomes can be attributed to teaching methods
Random assignment enables valid causal conclusions about teaching method effectiveness
Random assignment ensures both groups have similar distributions of prior knowledge, allowing causal attribution to teaching methods rather than pre-existing differences.
• Statistical Balance: Randomization balances all variables (known and unknown)
• Causal Validity: Only random assignment allows for causal inferences
• Probability Distribution: Groups are equivalent on average before treatment
Random Sampling: Selecting participants from the population using chance. Random Assignment: Allocating selected participants to experimental groups using chance.
Selecting participants from population using probability methods (ensures representativeness)
Allocating selected participants to treatment/control groups using chance (ensures equivalence)
Random sampling → generalizability; Random assignment → causation
Both are important but serve different inferential purposes
Random sampling ensures representative sample for generalization; random assignment ensures group equivalence for causal inference. Both are needed for strong experiments.
• Random Sampling: Enables population inferences (external validity)
• Random Assignment: Enables causal inferences (internal validity)
• Combined Power: Both together provide strongest experimental design
Assignment Probability: The likelihood that a participant will be assigned to a particular experimental group during random assignment.
Initially: 50 treatment spots available out of 100 total spots → P = 50/100 = 0.5 or 50%
After each assignment, available spots decrease in one group, but probability depends on remaining pool
Using complete randomization (all assignments determined simultaneously), each person has 50% chance
Proper randomization ensures equal probability regardless of assignment order
Each participant has a 50% probability of assignment to treatment group, maintained through proper randomization procedures.
• Equal Probability: All participants must have identical assignment chances
• Simultaneous Assignment: Determine all assignments at once to maintain probabilities
• Transparency: Document randomization method for verification
Blocking: A technique that divides participants into homogeneous subgroups (blocks) before random assignment to ensure balance on important variables.
Create two blocks: Block 1 (20 males), Block 2 (20 females)
Randomly assign 10 males to treatment, 10 to control; same for females
Treatment group: 10 males + 10 females; Control group: 10 males + 10 females
Both groups have identical gender distribution (50% male, 50% female)
Use blocking by gender: create male and female blocks, then randomly assign within each block to ensure balanced gender distribution in both groups.
• Block Creation: Group by important variables before randomization
• Within-Block Randomization: Maintain random assignment within each block
• Balance Achievement: Blocking guarantees balance on specified variables
Random Assignment: The process of allocating experimental units to treatment groups using chance, ensuring each unit has an equal probability of assignment to any group.
Confounding Variable: A variable that influences both the independent variable and the dependent variable, potentially creating spurious associations.
Internal Validity: The extent to which a study establishes a valid causal relationship between variables.
External Validity: The extent to which study results can be generalized to other populations or settings.
Blocking: A technique that groups similar experimental units together before random assignment to ensure balance on important variables.
- Preparation: List all participants and define treatment groups
- Randomization: Use computer software, random number tables, or other random mechanisms
- Implementation: Assign participants based on random sequence
- Documentation: Record assignment method and results for transparency
• Equal Probability: Every participant must have identical assignment probability
• Chance Determination: Assignment must be based solely on random mechanism
• Baseline Equivalence: Groups should be equivalent before treatment begins
• Causal Inference: Only random assignment enables valid causal conclusions
• Transparency: Assignment procedures must be documented and reproducible
Questions & Answers
Question: Why is random assignment necessary for causation but random sampling isn't? Can't we just use random sampling and get the same results?
Answer: These serve completely different purposes:
- Random Sampling: Ensures your sample represents the population (generalizability)
- Random Assignment: Ensures groups are equivalent before treatment (causation)
Key Point: You can have a perfectly random sample but still have systematic differences between groups if you don't randomize assignment. Random assignment is what creates the baseline equivalence needed for causal inference.
Example: Randomly sample 100 people, but assign all tall people to treatment and all short people to control - you have good sampling but no causal validity!
Question: What's the difference between blocking and stratification? Aren't they the same thing?
Answer: While similar in concept, they're used in different contexts:
- Blocking: Used in experiments during random assignment (before treatment)
- Stratification: Used in sampling to ensure representation from different groups
Process Similarity: Both involve dividing into homogeneous groups first, then applying randomization within groups.
Example: In an experiment, you might block by age (create age blocks, then randomize within each block). In sampling, you might stratify by age (ensure representation from each age group).
The key difference is timing: blocking happens during assignment, stratification during selection.
Question: How does random assignment handle variables we don't know about? How can it balance everything?
Answer: This is the beautiful mathematical property of randomization:
- Probability Theory: On average, random assignment balances ALL variables (known and unknown)
- Law of Large Numbers: As sample size increases, groups become more equivalent
- Statistical Guarantee: The expected difference between groups is zero for all variables
How It Works: Each participant has equal probability of assignment regardless of their characteristics. This means that any variable (age, gender, income, personality traits, genetic factors) gets distributed equally on average across groups.
It's like flipping coins - over many flips, heads and tails balance out. Similarly, random assignment balances all participant characteristics across groups.