UPHESC Assistant Professor Psychology 2025 PYQs

Concept: Each early school had one core assumption about mind and behaviour.

1. Structuralism (Wundt, Titchener) – break the mind into its elements by introspection → a.
2. Gestalt (Wertheimer, Köhler, Koffka) – experience depends on the pattern/whole → c.
3. Functionalism (James, Dewey, Angell) – mind and behaviour serve adaptation → d.
4. Behaviourism (Watson) – conditioned S–R responses are the building blocks → b.

Short trick: S-E (elements), G-P (pattern), F-A (adaptive), B-C (conditioning): a c d b.

Concept: Humanistic psychology (Maslow, Rogers) assumes people have free will and strive for self-actualisation.

1. Psychodynamic (Freud) – behaviour is determined by unconscious drives (psychic determinism).
2. Behaviourism – behaviour determined by environment/reinforcement.
3. Humanistic – the 'third force' stresses choice, responsibility and free will.

Short trick: Humanism = Human choice: free will belongs to the 'third force'.

Concept: The phenomenological approach (Rogers, Kelly, humanists) holds that behaviour is understood through the person's own subjective view of the world.

1. Psychodynamic – stresses unconscious conflicts, not conscious experience.
2. Behavioural – studies observable behaviour only.
3. Factor-analytic (Cattell, Eysenck) – derives traits statistically, an outside view.

Short trick: Phenomenon = what appears to ME: subjective experience.

Concept: When the criterion has two categories and predictors are continuous, use discriminant function analysis or binary logistic regression.

1. Discriminant analysis – finds the linear combination of predictors that best separates two (or more) groups.
2. Binary logistic regression – models the log-odds of a two-category outcome.
3. Multinomial logistic is meant for 3+ unordered categories; multivariate regression is for several continuous DVs.

Short trick: Two groups → 'Di'scriminant + 'Bi'nary logistic.

Concept: A predicted score $\hat{Y}$ from two or more predictors with an intercept and slopes is a multiple regression equation.

1. a = intercept; $b_1,b_2$ = partial regression coefficients.
2. Partial correlation gives a single r, not a prediction equation.
3. ANOVA compares means; SD is a dispersion index.

Short trick: One X → simple regression; many X's → multiple regression.

Concept: SD is the square root of the mean of squared deviations taken from the arithmetic mean.

1. $\sigma =\sqrt{\frac{\sum (X-\overline{X}{)}^2}{N}}$ – every deviation is measured from the mean.
2. So SD tells how widely scores scatter around the mean.
3. Lowest/highest values define the range, not SD.

Short trick: SD = 'Spread from the Mean'.

Concept: In a repeated-measures design the same subjects serve in all conditions, so individual differences are removed from the error term.

1. Between-subject variability is partitioned out → smaller error variance.
2. Hence greater statistical power with fewer subjects.
3. It is not automatically more valid – order/carry-over effects threaten it, so counterbalancing is needed.

Short trick: Same people in every condition = each is his own control → less error.

Concept: In probability sampling every unit has a known, non-zero chance of selection.

1. Simple random, stratified random and cluster/area sampling all use random selection.
2. Quota sampling fills fixed quotas by the investigator's convenience/judgement – no random selection.
3. Other non-probability types: purposive, convenience, snowball.

Short trick: Quota = Quick and non-random.

Concept: An interval scale has equal units, so differences between scores are meaningful, but its zero is arbitrary.

1. Categories only → nominal scale.
2. Ordering only → ordinal scale (interval also orders, but that is not its defining feature).
3. True/meaningful zero → ratio scale.
4. Equal intervals so that size of difference shows magnitude → interval scale.

Short trick: NOIR: Nominal-name, Ordinal-order, Interval-equal gaps, Ratio-real zero.

Concept: Statistics exists precisely to describe and explain variability, not to ignore it.

1. Descriptive statistics summarise data from many sources.
2. Inferential statistics generalise from sample to population.
3. Multivariate methods (factor analysis etc.) reduce and explain large data sets.
4. Ignoring variability defeats the purpose of statistics.

Short trick: Statistics = science of variation.

Concept: The Mann-Whitney U test is the non-parametric counterpart of the independent t-test with about 95% power-efficiency.

1. Mann-Whitney U uses ranks of two independent groups – asymptotic relative efficiency ≈ 0.95 vs t.
2. Median test uses only above/below-median information → much less powerful (≈ 0.64).
3. Chi-square is for frequencies; Friedman is for related samples (k conditions).

Short trick: Two independent groups: t-test ↔ Mann-Whitney U (95% as powerful).

Concept: KMO and Bartlett's test check whether the data are suitable for factor analysis; they do not decide how many factors to keep.

1. Latent root (Kaiser) criterion – keep factors with eigenvalue > 1.
2. Scree plot (Cattell) – keep factors above the 'elbow'.
3. Velicer's MAP (minimum average partial) – a retention rule.
4. KMO (sampling adequacy) and Bartlett's sphericity – pre-analysis suitability tests.

Short trick: KMO-Bartlett = 'Can I factor?'; eigenvalue/scree/MAP = 'How many factors?'

Concept: Kruskal-Wallis H compares three or more independent groups on ranks – the rank-based version of one-way ANOVA.

1. One factor, k independent groups → one-way ANOVA / Kruskal-Wallis.
2. Repeated measures with k conditions → Friedman test (non-parametric two-way by ranks).
3. Factorial/ANCOVA have no single standard rank equivalent here.

Short trick: KW = k groups, one way; Friedman = same people, k conditions.

Concept: Mann-Whitney U is a rank test, so it needs data at least on an ordinal scale.

1. Scores of two independent groups are pooled and ranked; U is computed from rank sums.
2. Interval data with normality would call for the t-test instead.
3. Nominal/dichotomous frequency data → chi-square.

Short trick: Rank tests (U, H, Wilcoxon, Friedman, rho) = ordinal data.

Concept: Univariate analysis describes one variable at a time; correlation needs two variables.

1. Frequencies, percentages, mean/median/mode, SD – all describe a single variable.
2. Correlation shows the relationship between two variables → bivariate analysis.

Short trick: Correlation is a 'co-' measure – it always needs a partner variable.

Concept: Classical test theory: X = T + E, where E (obtained minus true score) is random measurement error.

1. E is assumed random: mean zero and uncorrelated with T.
2. Systematic error shifts all scores the same way and is absorbed into the 'true' score – it affects validity, not reliability.
3. Shared/unique variance are factor-analytic terms.

Short trick: X – T = E (random error); reliability = proportion of variance free of it.

Concept: A high item-total correlation means the item separates high and low scorers and measures the same thing as the whole test.

1. Discrimination: point-biserial item-total r is a standard discrimination index.
2. Homogeneity/internal consistency: items correlating with the total measure the same trait.
3. Difficulty is the proportion passing (p), not a correlation.

Short trick: Item-total r = D-index + homogeneity; p-value = difficulty.

Concept: Factor analysis checks whether a test's items group into the theoretical dimensions of the construct – factorial (construct) validity.

1. Content/face validity – judged by experts/appearance, no statistics.
2. Empirical (criterion) validity – correlation with an external criterion.
3. Construct validity – factor analysis, convergent–discriminant (MTMM) evidence.

Short trick: Factor analysis → factorial validity → a type of construct validity.

Concept: Visualization (guided imagery) is deliberately picturing a calm, pleasant scene to reduce stress.

1. Concentrative meditation focuses on one object, sound (mantra) or the breath.
2. Progressive muscular relaxation tenses and relaxes muscle groups.
3. Imagining a quiet beach = mental imagery → visualization.

Short trick: 'Envision' = visualization.

Concept: Splitting one test into odd and even halves and correlating them gives split-half (internal consistency) reliability.

1. The half-test r is stepped up with Spearman–Brown: $r_{tt}=\frac{2r_{hh}}{1+r_{hh}}$.
2. Test-retest – same test given twice; parallel form – two equivalent forms.

Short trick: Odd–even = one test, one sitting, two halves → split-half.