Explain the use of parallel tests in psychological studies.
5 marks · directive: Explain
The core of a good answer
Parallel tests are two forms of a test that measure the same true score with equal error variance, so they have equal means, variances and intercorrelations with other measures; the correlation between them is the reliability coefficient ρ = σ_T²/σ_X², and they are used to estimate reliability (parallel or alternate forms method), to retest without practice or memory effects in pre-post studies, to lengthen tests (Spearman-Brown) and to correct correlations for attenuation.
An outline written by upsc.help, not an official UPSC answer; every fact checked by a separate review. How we work
What the question asks
Define parallel tests in classical test theory, then explain their uses with formulas where relevant. Short answer.
Answer outline
Introduction
Classical test theory: observed score X = T + E, with E(E) = 0 and E uncorrelated with T.
Definition
- Tests X and X′ are parallel if X = T + E and X′ = T + E′ with the same true score T and Var(E) = Var(E′).
- Then means and variances of X and X′ are equal, and both correlate equally with any third variable.
Uses
- Reliability: ρ_XX′ = Cov(X, X′)/(σ_X σ_X′) = σ_T²/σ_X², the proportion of observed variance that is true variance.
- Alternate-forms testing: one form before and the other after training or treatment avoids memory and practice effects of repeating the same items.
- Standard error of measurement σ_X√(1 − ρ_XX′) gives confidence bands for individual scores.
- Test length: Spearman-Brown formula ρ_kk = kρ/[1 + (k − 1)ρ] predicts the reliability of a test lengthened k times with parallel parts; split-half reliability treats the halves as parallel.
- Correction for attenuation: ρ_TxTy = ρ_xy/√(ρ_xx′ρ_yy′).
Conclusion
Parallel forms are the theoretical basis of reliability in classical test theory and a practical device for repeated testing without contamination.
Examples and data to cite
- Spearman and Brown published the prediction formula independently in 1910, in the British Journal of Psychology.
Facts to remember
- X = T + E; parallel forms share T and have equal error variance.
- Reliability ρ = σ_T²/σ_X² = correlation between parallel forms.
- Spearman-Brown: ρ_kk = kρ/[1 + (k − 1)ρ].
- Standard error of measurement = σ_X√(1 − ρ).