Internal Validity: The extent to which a study establishes a valid causal relationship between variables, free from confounding factors.
- Selection Bias: Students who choose tutoring may be more motivated
- Confounding Variables: Socioeconomic status, parental involvement
- Self-Selection: Motivated students choose tutoring
- History Effects: Other interventions during study period
No random assignment allows systematic differences between groups
Motivation, socioeconomic status, parental support, academic ability
Cannot conclude tutoring caused grade improvement due to confounding
Use random assignment or statistical controls for confounding variables
Major threats include selection bias and confounding variables. Without random assignment, cannot conclude tutoring caused improved grades. Need random assignment or statistical controls.
• Random Assignment: Primary method for eliminating confounding variables
• Confounding Control: Balance known and unknown confounding variables
• Causal Inference: Only experiments with proper controls allow causal conclusions
External Validity: The extent to which study results can be generalized to other populations, settings, and conditions.
Elite private school students may have different resources, motivation, and abilities
Resources, socioeconomic status, teacher quality, student motivation
Results may not apply to public schools or different socioeconomic contexts
External validity severely limited by non-representative sample
External validity is severely limited because elite private school students differ significantly from general population in resources, motivation, and other factors.
• Representativeness: Sample must reflect target population characteristics
• Context Matters: Environmental and situational factors affect generalizability
• Demographic Matching: Key characteristics should match target population
Response Bias: Systematic error in survey responses due to question wording, social desirability, or other factors that influence truthful reporting.
"Excellent teachers" primes positive response
People may agree to appear supportive of education
Results likely inflated due to question bias
Use neutral, open-ended questions to reduce bias
Leading question and social desirability bias invalidate results. The 95% agreement likely overestimates true support due to question wording effects.
• Neutral Wording: Questions should be unbiased and neutral
• Leading Language: Avoid loaded or emotionally charged terms
• Validation: Biased questions produce invalid results regardless of sample size