Saral Shiksha Yojna
Courses/Behavioral Research: Statistical Methods

Behavioral Research: Statistical Methods

CG3.402
Vinoo Alluri•Monsoon 2025-26•4 credits
Sample Papers/Mock Paper 3 — All Units Mix

Mock Paper 3 — All Units Mix

Duration: 120 min • Max marks: 100

Section A — MCQ (20 × 1 = 20)

20 marks
  1. 1.True for a paired t-test: (a) Two independent samples (b) Two related samples (pre/post on same subjects) (c) 3+ groups (d) Categorical DV1 m
  2. 2.MS_between / MS_within in ANOVA represents: (a) Effect size (b) F-statistic (c) Variance ratio (d) Both (b) and (c)1 m
  3. 3.In Bayesian terminology, 'evidence' P(D) acts as: (a) Prior (b) Normalising constant (c) Posterior (d) Conditional1 m
  4. 4.Order of sample-size effects on the t-distribution: (a) Smaller n → t closer to Normal (b) Larger n → t with heavier tails (c) Larger n → t approaches Normal (d) n doesn't affect t1 m
  5. 5.Test with sensitivity 0.99 and specificity 0.99, prevalence 0.1%. P(disease | positive) ≈ ?1 m
  6. 6.Sign of overfitting in regression: (a) High training R², much lower test R² (b) Both high (c) Both low (d) Linear relationship1 m
  7. 7.'P(this specific 95% CI contains μ) = 95%' is: (a) Correct frequentist (b) Common misinterpretation; procedure has 95% coverage (c) Bayesian credible-interval interpretation (d) Both (b) and (c)1 m
  8. 8.Pearson r requires which scale? (a) Nominal (b) Ordinal (c) Interval / ratio (d) Any1 m
  9. 9.Stratified sampling is most useful when: (a) No subgroups (b) Subgroups differ on the outcome and proportions matter (c) Data collected randomly (d) Convenience sample wanted1 m
  10. 10.An ANOVA omnibus F is non-significant. You should: (a) Run pairwise post-hocs anyway (b) Stop and conclude no group differences (c) Increase α (d) Switch to χ²1 m
  11. 11.'Ecological validity' refers to: (a) Internal consistency (b) Reliability over time (c) Generalisability to real-world settings (d) Statistical power1 m
  12. 12.Binomial(n=20, p=0.5) has mean and variance: (a) 10 and 5 (b) 20 and 10 (c) 10 and 10 (d) 5 and 51 m
  13. 13.Comparing two correlated proportions (same people, two yes/no items) — appropriate test: (a) χ² independence (b) McNemar's test (c) Pearson r (d) ANOVA1 m
  14. 14.Bias-variance trade-off: (a) Bias and variance always equal (b) Reducing one tends to increase the other (c) Both can be eliminated together (d) Bias only applies to means1 m
  15. 15.Spider/radar plots are generally considered: (a) Highly informative (b) Often misleading due to area dependence on axis order (c) Standard for ANOVA (d) The only multivariate plot1 m
  16. 16.Correct order from least to most informative scale: (a) Interval, Ratio, Ordinal, Nominal (b) Nominal, Ordinal, Interval, Ratio (c) Ordinal, Nominal, Ratio, Interval (d) Ratio, Interval, Ordinal, Nominal1 m
  17. 17.Funnel-shape residual plot indicates: (a) Linearity (b) Normality (c) Heteroscedasticity (d) Multicollinearity1 m
  18. 18.Cronbach's α is a measure of: (a) Type I error (b) Internal consistency reliability (c) Construct validity (d) Effect size1 m
  19. 19.Non-parametric correlation: (a) Pearson r (b) Spearman ρ (c) Cohen's d (d) η²1 m
  20. 20.Power analyses are typically performed: (a) After data collection (b) Before data collection to plan n (c) Only in Bayesian analysis (d) To compute p-values1 m

Section B — MSQ (10 × 2 = 20)

20 marks
  1. 1.Correct about ANOVA: (a) F is a ratio of variances (b) ANOVA tests whether all means are equal (c) ANOVA assumes equal variances (d) Significant F identifies which groups differ (e) Effect size is η²2 m
  2. 2.Useful for visualising distributions: (a) Boxplot (b) Histogram (c) Density plot (d) Pie chart (e) Raincloud plot2 m
  3. 3.Valid concerns about p = 0.049: (a) Below 0.05 so automatically valid (b) Could be p-hacking (c) Effect size might be tiny (d) Power might be low (e) Significance ≠ practical importance2 m
  4. 4.Types of validity: (a) Internal (b) Inter-rater (c) External (d) Construct (e) Ecological2 m
  5. 5.Conditions supporting causal inference: (a) Random assignment (b) Temporal precedence (c) Strong correlation (d) Ruling out confounds (e) Replicability across contexts2 m
  6. 6.Apply to χ² tests: (a) For categorical data (b) (O−E)²/E summed over cells (c) df based on cells (d) Expected cell counts ≥ 5 (e) Effect size Cramér's V2 m
  7. 7.Apply to Bayes Factor: (a) Can provide evidence for H₀ (b) Depends on prior (c) Robust to optional stopping (d) Always > 1 (e) Quantifies evidence continuously2 m
  8. 8.Advantages of within-subjects designs: (a) Higher power (b) Fewer participants needed (c) No carryover (d) Controls individual differences (e) Allows time-irreversible IVs2 m
  9. 9.Flagged as outliers by Tukey's 1.5×IQR rule: (a) Q1 − 2.0×IQR (b) Q3 + 1.6×IQR (c) Q1 + 0.5×IQR (d) Median (e) Q3 + 3×IQR2 m
  10. 10.Strengthen external validity: (a) Diverse sample (b) Multiple settings (c) Strict lab control (d) Replication (e) Convenience sampling of college sophomores2 m

Section C — Short descriptive (6 × 5 = 30)

30 marks
  1. 1.Bessel's correction (n−1 in sample variance) — why it makes the estimator unbiased.5 m
  2. 2.One-tailed p = 0.03 supports a directional hypothesis. Corresponding two-tailed p? When is each appropriate?5 m
  3. 3.Why should effect size always be reported alongside p-values?5 m
  4. 4.Briefly describe simple random, stratified, convenience, and snowball sampling. Rank by typical external validity.5 m
  5. 5.Define p-hacking. Three specific behaviours and one antidote.5 m
  6. 6.Define a confidence interval and a credible interval; explain how interpretations differ.5 m

Section D — Long descriptive (3 × 10 = 30)

30 marks
  1. 1.Walk through the construction and interpretation of a 95% CI for a population mean μ. (i) Formula + assumptions. (ii) Behaviour with n, confidence level, σ. (iii) CI vs prediction interval. (iv) Critique 'there is a 95% chance the true mean is between 4.2 and 5.8.'10 m
  2. 2.Discuss randomisation in experimental design. (i) What it accomplishes. (ii) Random assignment vs random sampling. (iii) Alternatives when randomisation is impossible. (iv) Limitations of randomised experiments.10 m
  3. 3.20 Likert items measure 'social anxiety'. (a) Check reliability. (b) Check construct(s). (c) Reduce to scale scores for downstream analyses. Describe the full factor-analytic pipeline.10 m

Track your attempt locally — score and time are recorded in your browser. (Coming soon: timed-attempt mode.)