Research methods p2

Mean AO1

Definition: Calculates the average score by dividing the total sum of values by the number of values, Example: Values: 4, 6, 7, 9 → Total: 26 → Divide by 4 → Mean = 6.5, Strengths: Most sensitive as it includes all scores, Provides a representative and reliable measure, Limitations: Sensitive to extreme values (outliers), Mean may not exist within the original data set (e.g., 6.5 doesn't appear)

Median AO1

Definition: Positional average, calculates the middle value in ordered data, Example: For an odd set: 20, 43, 48, 56, 67, 78, 92 → Median = 56, For an even set: 15, 16, 18, 19, 22, 24 → Median = 18.5 (average of middle values: 18 & 19), Strengths: Not affected by extreme values, Suitable for data sets with anomalous scores or qualitative ranking, Easy to calculate once data arranged from lowest to highest order, Limitations: Does not represent all data (less reliable), Time-consuming for large data sets, Extreme higher and lower values may be ignored

Mode AO1

Definition: Identifies the most frequent score(s) in a data set, Some datasets may have two modes (bi-modal), many modes (multi-modal), or no mode if all scores differ, Example: Data: 3, 3, 3, 4, 4, 5, 6, 6, 6, 6, 7, 8 → Mode = 6, Strengths: Unaffected by extreme values, Useful for analysing frequency in qualitative data, Very easy to calculate, Limitations: Multiple modes can blur data interpretation, Likely unrepresentative in small data sets

Range AO1

Measures of dispersion calculate the spread of scores and how much they vary from the mean or median, Low dispersion = scores cluster around central tendency, High dispersion = scores spread apart, Identical scores = dispersion = 0, Mean, median, mode identical, Two measures of dispersion: Range, Standard deviation, Range describes difference between lowest and highest scores, Provides information on the gap between highest and lowest scores, Calculation: Subtract lowest from highest, e.g., 4, 4, 6, 7, 9 → 9-4 = 5, Add 1 if rounding applied → Range = 6

Range AO3

Strengths: Provides a broad overview of data, Simple and easy to calculate, Limitations: Provides no information on other scores, Lacks validity as it does not indicate degree of variation from mean, Not very stable or representative

Standard Deviation AO1

Calculates how scores deviate from the mean, Provides insight into clustering/spread, Low SD = scores tightly clustered → higher reliability, High SD = scores spread out → lower reliability, Normal distributions have low SD, Calculation steps: Calculate mean, Subtract mean from each score, Square the differences, Sum squared differences, Divide by n-1, Square root of variance

Standard Deviation AO3

Strengths: Indicates how scores are distributed, Shows reliability and consistency, More sensitive than range → more valid, Limitations: Time-consuming, Can be skewed by extreme outliers

Sign Test AO1

Steps: Calculate difference (subtract second condition score from first), Assign a sign (+/-), Calculate N (number of participants with a difference), Identify less frequent sign (S) → calculated value, Find critical value from table, Compare calculated and critical values, Significance: Calculated ≤ critical → significant, reject null, Calculated > critical → not significant, accept null

Significance Levels AO1

1% value = stringent benchmark, avoids type 1 error, 5% value = more lenient benchmark, balances risk of type 1 and type 2 errors

Probability & Significance AO1

Probability = extent to which something is likely to happen, Expressed 0–1, 0 = impossible, 1 = certain, Calculated as desired outcomes / total outcomes, Significance: p < 0.05 → result unlikely due to chance, p < 0.01 → stringent, used for human cost or contradictory evidence, Critical values used to compare calculated statistic

Type I & II Errors AO1

Type I: Reject null when should accept → false positive, More likely if p too high, Type II: Accept null when should reject → false negative, More likely if p too low, Using p=0.05 guards against both

Normal Distribution AO1

Data spread around mean, Symmetrical, bell-shaped, Most scores near mean, Extreme scores in tails, Tails never touch x-axis, Examples: height, weight, shoe size, Used to identify deviance e.g., IQ ± 2 SD, postpartum depression score, empathy score

Skewed Distributions AO1

Asymmetrical, one tail longer than other, Mean most affected, Positive skew: most values left, long right tail, Examples: age of first job, difficult test scores, Negative skew: most values right, long left tail, Examples: age of retirement, easy test scores

Reliability – Definition & Types

Reliability = consistency of a measure or study, Internal reliability = how consistent parts of a test are with each other, External reliability = consistency over time or between situations, Inter-rater reliability = agreement between observers (should be r ≥ 0.8), Reliable data are replicable and consistent, increasing scientific credibility

Reliability – Testing & Improving

Improve reliability by using clear operational definitions, standardised procedures, pilot studies, and quantitative data (scores, reaction times)

Test–retest (external reliability)

• Same participants complete the test/questionnaire at different times.

• Consistent results across time = reliable.

• Split-half (internal reliability)

  • Questionnaire is divided into two halves.

  • If responses to both halves are similar, the test is internally reliable.

• Inter-observer reliability

  • Two (or more) observers record behaviour independently for the same person/condition

  • Their data sets are compared (often via correlation).

  • A strong positive correlation = reliable behavioural categories.

Validity – Definition & Types

Validity = accuracy — whether the research accurately measures what it intends to, Internal validity = results due to IV not extraneous variables, External validity = can findings be generalised (ecological, population, temporal), Face validity = procedure /questions looks like it measures what it should, Construct validity = truly measures the intended concept, Concurrent validity = correlates with similar tests, Predictive validity = predicts future behaviour

Validity – Improving Validity

To improve internal validity → control variables, standardise procedures, use blinding and random allocation, To improve external validity → realistic settings, representative samples, replicate in different contexts, In observations → use covert methods and clear categories, In questionnaires → use lie scales and reverse-worded items to reduce bias

FOS – Theory Construction & Hypothesis Testing

Theories = general explanations built from evidence, Hypotheses = testable predictions from theories, Two types: directional (predicts direction) and non-directional (predicts difference only), Uses the hypothetico-deductive model → theory → hypothesis → test → refine, Replication checks reliability and supports/refines theory, Statistical testing (e.g., p<0.05) decides whether to accept or reject the null

FOS – Falsifiability

Popper: scientific theories must be testable and capable of being proven wrong, Falsifiability distinguishes science from pseudoscience, e.g., behaviourism and cognitive psychology are testable, psychodynamic ideas (unconscious) are not, Example: Maguire et al. (2000) → measurable hippocampal size → falsifiable, Encourages scientific rigour but hard to apply to abstract behaviours

FOS – Replicability

Replicability = repeating research to check results are consistent showing method has scientific credibility, Needs standardised procedures and clear operationalisation, Quantitative data → more replicable, Qualitative → harder to reproduce, The replication crisis in psychology shows why open data and transparency matter, consistency within replication strengthens theories and credibility as it is less likely that the findings are due to chance alone, improving reliability.

FOS – Objectivity

Objectivity = data collected without bias or personal interpretation, Achieved through lab settings, standardised measures, and quantitative data, Reduced by subjective methods (e.g., case studies, interviews), Researcher bias can be minimised using blinding and inter-rater reliability, Peer review helps maintain objectivity and scientific quality

FOS – Paradigms & Paradigm Shifts

Kuhn: a paradigm = shared assumptions and methods in a science, Psychology has competing paradigms (biological, cognitive, behaviourist) → seen as pre-paradigmatic, Paradigm shift = major change when old model replaced, e.g., behaviourism → cognitive revolution (1950s), Shows psychology’s progress toward scientific maturity

FOS – Peer Review

Peer review = experts check research before publication, Ensures quality, validity, originality, and ethics, Prevents flawed or fraudulent work, Uses anonymous reviewers to reduce bias, Weaknesses = bias, suppression of new ideas, publication bias for positive results, Strength = improves scientific credibility of published work

FOS – Psychology as a Science

For = empirical methods, hypothesis testing, control, quantifiable data, replicability, integration with biology, Against = human behaviour complex, some approaches untestable (psychodynamic), qualitative data subjective, case studies idiographic, Overall = psychology is becoming more scientific through replication, brain imaging, and cognitive neuroscience

concurrent validity

high concurrent validity is a close agreement between the scores of a test compared to an already established test (e.g. a school assessment compared to a standard IQ test) if the correlation coefficient between the two testa is +0.8 is is said to have high concurrent validity.

checking reliability vs checking validity

Correlation Coefficient (r value) - reliability

• Measures the strength and direction of the relationship between two sets of scores (e.g., Observer A vs Observer B).

• Ranges from –1 to +1.

  • +1 = perfect positive correlation.

  • 0 = no correlation.

  • –1 = perfect negative correlation.

• In psychology, r ≥ +0.8 is usually considered strong enough to claim good inter-observer reliability.

How p values can relate to validity = p=<0.05

1. Concurrent validity

   • If a new test correlates strongly with an established valid test, you can calculate a correlation coefficient and its p value.

   • A significant p value here supports the claim that the correlation is real, which indirectly supports validity

Predictive validity

• If a test is meant to predict future outcomes (e.g., exam performance), researchers can test whether scores significantly predict those outcomes.

• A significant p value here supports predictive validity.