← Educational Leadership & Research Resources A Complete Methods Resource

Quantitative Inquiry in Education,
from measurement to defensible inference.

A rigorous, interactive guide to measuring, comparing, and predicting educational phenomena — the deductive logic beneath the numbers, how variables become measurable, how samples support claims about populations, and how to choose and interpret the right statistical test. Built for graduate researchers, evaluators, and educational leaders.

4
Levels of measurement
4
Probability sampling types
5
Core statistical tests
.05
The conventional α
Memorandum
To Educational leaders, researchers & decision-makers
From Freddrick Logan — Technologist & Education
Date 2026
Re Quantitative Inquiry in Education

A rigorous, interactive field guide that carries an educational researcher from measurement to defensible statistical inference.

  • Grounds the deductive logic of quantitative inquiry — how constructs become measurable variables and how sample data support claims about a population.
  • Explains hypothesis testing end to end: null and alternative hypotheses, α, p-values, Type I and Type II errors, and the drivers of statistical power.
  • Distinguishes statistical significance from practical significance with a live slider showing how effect size (Cohen's d) stays fixed while p shrinks with sample size.
  • Provides an interactive test selector and reference table that map an analytic goal to the right test — from t-tests and ANOVA to regression and chi-square.
  • Sharpens critical appraisal of results — p-hacking, multiple comparisons, and the replication crisis — through a spot-the-flaw quiz and a defensible-reporting checklist.
  • Frames reliability, validity, and research ethics as the conditions for trustworthy, transparent, and reproducible educational research.
Foundations

The logic of quantitative inquiry

Quantitative research reasons deductively: it begins with theory, derives testable hypotheses, operationalizes concepts as measurable variables, and uses data from a sample to draw defensible inferences about a population. It asks how much, how many, and how related.

From concept to measurement

  • Constructs — abstract ideas (motivation, self-efficacy, achievement)
  • Operationalization — the rules that turn a construct into a measure (a test score, a Likert scale)
  • Variables — the measured attributes that vary across cases
  • Data — the numeric values recorded for each unit of analysis
Operationalization is where rigor begins. A construct is only as valid as the operations used to measure it.

Hypotheses & inference

TermMeaning
Null (H₀)No effect / no difference / no relationship
Alternative (H₁)A specified effect, difference, or relationship exists
PopulationThe full group a claim is about
SampleThe observed subset used to estimate it
ParameterA true population value (μ, σ, ρ)
StatisticAn estimate from the sample (x̄, s, r)

Inferential testing asks: if H₀ were true, how surprising is our sample result?

Directional vs. non-directional. A one-tailed (directional) hypothesis predicts the sign of an effect ("the treatment raises scores"); a two-tailed test only predicts a difference. Choose the tail before seeing the data.
Inference

Hypothesis testing, fully

Null hypothesis significance testing is a proof-by-contradiction. You assume the null is true, ask how surprising your data would be under that assumption, and reject the null only when the data are surprising enough. Rejecting the null never proves the alternative — it only says the null is an implausible explanation for what you saw.

The logic of rejection

  • State H₀ and H₁. The null (H₀) posits no effect, no difference, no relationship; the alternative (H₁) is what you actually suspect.
  • Set α before collecting data. α is the risk of a false positive you are willing to tolerate — conventionally .05, sometimes .01.
  • Compute a test statistic and its p-value. The p-value is the probability of data at least this extreme if H₀ were true.
  • Decide. If p < α, reject H₀ ("statistically significant"); otherwise fail to reject it.
  • Never "accept" H₀. Failing to reject is not proof of no effect — it may simply reflect low power.
Fail to reject ≠ prove the null. Absence of evidence is not evidence of absence. A non-significant result with a wide confidence interval is inconclusive, not a confirmation of "no effect."

The two ways to be wrong

Every decision faces two realities — the null is really true, or it is really false — crossed with two verdicts. Two of the four cells are errors.

Decision \ RealityH₀ is trueH₀ is false
Reject H₀Type I error
False positive (prob = α)
Correct
True positive (prob = 1 − β = power)
Fail to reject H₀Correct
True negative (prob = 1 − α)
Type II error
False negative (prob = β)

Lowering α guards against Type I errors but, all else equal, raises β and lowers power. The two risks trade off.

Statistical power — and what drives it

Definition

Power = 1 − β

The probability of correctly rejecting a false null — of detecting a real effect when it exists. A common target is .80 (Cohen, 1988).

Sample size

Bigger n → more power

Larger samples shrink the standard error, making true effects easier to detect. This is the lever the researcher controls most directly.

Effect size

Bigger effect → more power

Large effects are easy to detect; small ones demand large samples. Power analysis must assume a plausible effect size up front.

α & variance

Looser α, less noise → more power

A larger α or a more reliable, less variable measure both raise power. A one-tailed test has more power than a two-tailed test in the predicted direction.

Do the power analysis first. An a priori power analysis fixes the sample size needed to detect the smallest effect worth caring about. Underpowered studies both miss real effects and — paradoxically — inflate the size of any effect they do detect (the "winner's curse").
Measurement

Levels of measurement & variable types

Stevens' four levels determine which statistics are legitimate. Mistaking one for another — averaging ordinal ranks, for instance — produces numbers that look precise but mean little.

Nominal

Labels, no order

Categories with no rank. Examples: school, program, gender, treatment vs. control.

Allowed: counts, mode, chi-square

Ordinal

Ordered, unequal gaps

Ranked, but distances aren't equal. Examples: Likert agreement, class rank, letter grades.

Allowed: median, percentiles, rank tests

Interval

Equal gaps, no true zero

Equal intervals; zero is arbitrary. Examples: temperature (°C), many scaled test scores.

Allowed: mean, SD, correlation, t / ANOVA

Ratio

Equal gaps, true zero

A meaningful zero allows ratios. Examples: age, count correct, minutes on task.

Allowed: all of the above + ratios

Roles a variable can play

  • Independent (IV) — the presumed cause or predictor, manipulated or grouped
  • Dependent (DV) — the outcome measured for its response
  • Confounding variable — an unmeasured third variable that distorts the IV–DV relationship
  • Mediator — carries the effect of the IV to the DV
  • Moderator — changes the strength or direction of the effect
  • Control variable — held constant to isolate the effect of interest

How outcomes are distributed

  • Categorical — values fall into groups (nominal or ordinal)
  • Continuous — any value within a range (interval or ratio)
  • Discrete — countable whole numbers (number of absences)
  • Binary / dichotomous — two categories (pass/fail, retained/promoted)
Why it matters: the outcome's type is the single biggest driver of which test is appropriate.
Measurement quality

Reliability & validity of measures

Reliability is consistency; validity is accuracy. A bathroom scale that reads five pounds heavy every time is perfectly reliable yet invalid. The reverse is impossible: a measure cannot be valid without first being reliable — reliability is a necessary but not sufficient condition for validity.

Reliability

Four ways to gauge consistency

TypeWhat it checksHow
Test–retestStability over timeCorrelate scores from the same people on two occasions
Parallel / alternate formsEquivalence of two versionsCorrelate scores on two equivalent forms of the instrument
Internal consistencyItems measure one constructCronbach's α (or split-half, KR-20); α ≥ .70 often acceptable
Inter-raterAgreement between observersCohen's κ or intraclass correlation (ICC) across raters
Validity

Does it measure the right thing?

TypeQuestion it answersEvidence
ContentDo items cover the whole domain?Expert judgment; a blueprint mapping items to the construct
Criterion — predictiveDoes it forecast a future outcome?SAT scores predicting later college GPA
Criterion — concurrentDoes it agree with a present standard?A new screener vs. an established diagnostic given at the same time
ConstructDoes it behave like the theory says?Convergent & discriminant evidence; factor analysis
Cronbach's α, read carefully. A high α signals items hang together, but it rises mechanically with more items and does not by itself prove unidimensionality or validity. Report it per scale, not for a whole questionnaire of mixed constructs (Nunnally & Bernstein, 1994).
Sampling & design

How the study is built

Probability sampling supports generalization to a population; the design determines what kind of claim — causal or associational — the data can bear. Select a design to see its logic.

Probability sampling

Simple random

Equal chance for all

Every member has the same probability of selection. The benchmark against which others are judged.

Stratified

Sample within strata

Divide the population into subgroups (grade, SES), then sample within each — improves precision and representation.

Cluster

Sample whole groups

Randomly select intact clusters (schools, classrooms), then study all within. Efficient but less precise.

Systematic

Every k-th case

Choose a random start, then take every k-th unit from an ordered list. Simple, but beware periodicity in the list.

True experimental

Manipulates the IV and uses random assignment to conditions, so groups are equivalent in expectation. This is the strongest warrant for a causal claim.

Best question: "Does X cause a change in Y?"

Randomized controlled trialPretest–posttest control group
  • Key featureRandom assignment + manipulation
  • WarrantCausal inference
  • StrengthHigh internal validity
  • Trade-offFeasibility & ethics in schools

Quasi-experimental

Manipulates the IV but without random assignment — using intact groups (existing classrooms). Causal claims are weaker and must address selection threats.

Best question: "Does X plausibly cause Y when randomization isn't possible?"

Nonequivalent groupsInterrupted time seriesRegression discontinuity
  • Key featureManipulation, no randomization
  • WarrantConditional causal inference
  • ThreatSelection bias between groups
  • Common inApplied educational settings

Correlational

Measures variables as they naturally occur and quantifies their association — no manipulation. Establishes relationship and prediction, not causation.

Best question: "How, and how strongly, are X and Y related?"

Correlation is not causationPrediction studies
  • Key featureNo manipulation, measured as-is
  • WarrantAssociation & prediction
  • StatisticsPearson r, regression
  • CautionThird-variable / directionality problems

Descriptive / survey

Describes the characteristics of a population or phenomenon as it exists — frequencies, distributions, trends — without testing relationships between manipulated variables.

Best question: "What is the current state of ___?"

Cross-sectional surveyLongitudinal / trend
  • Key featureDescribes, does not explain
  • WarrantPrevalence & description
  • StatisticsDescriptive summaries
  • Common toolValidated survey instruments

Validity threats to guard against

TypeThreatWhat it undermines
InternalHistoryAn outside event, not the IV, caused the change
InternalMaturationParticipants change naturally over time
InternalSelectionGroups differed before treatment
InternalTesting & instrumentationThe measure itself changes scores
InternalRegression to the meanExtreme scorers drift toward average
InternalAttrition (mortality)Non-random dropout biases groups
ExternalPopulation validityResults may not generalize beyond the sample
ExternalEcological validityResults may not hold in other settings/times
The core tension. Tightly controlled designs maximize internal validity but can sacrifice external validity — and vice versa. Naming the trade-off is part of a defensible design (Shadish, Cook & Campbell).
Causal reasoning

Correlation is not causation

Two variables can move together for reasons that have nothing to do with one causing the other. Before an association can be read as a cause, it has to survive three classic tests — and even then, only a well-designed experiment closes the gap decisively.

Criterion 1

Covariation

The presumed cause and effect must actually be related — as X changes, Y changes. This is the part a correlation establishes; it is necessary but far from sufficient.

Criterion 2

Temporal precedence

The cause must come before the effect in time. Correlational and cross-sectional data usually cannot establish which came first — the directionality problem.

Criterion 3

No plausible alternative

Rival explanations — especially confounders — must be ruled out. This is where observational designs are most vulnerable (Shadish, Cook & Campbell, 2002).

The confounding (third-variable) problem

A confounder is a variable associated with both the presumed cause and the outcome, manufacturing a correlation between them. Ice-cream sales correlate with drowning deaths — but summer heat drives both; neither causes the other.

  • Reverse causation: maybe Y causes X, not X causes Y.
  • Common cause: a lurking Z drives both X and Y.
  • Selection: the groups you compared already differed.

What experiments buy you

Random assignment is the engine of causal inference. By allocating participants to conditions by chance, it makes the groups equivalent in expectation on every variable — measured or not, known or unknown — so any post-treatment difference can be attributed to the manipulation.

  • Neutralizes confounders you never even thought to measure.
  • Establishes temporal precedence by design (treatment precedes outcome).
  • Turns "associated with" into "causes," within the study's context.
When you can't randomize, quasi-experimental designs (regression discontinuity, difference-in-differences, matching, instrumental variables) approximate a counterfactual — but each rests on assumptions that must be argued, not assumed.
Analysis

Descriptive vs. inferential statistics

Descriptive statistics summarize the data in hand. Inferential statistics use the sample to reason — under uncertainty — about the population that produced it.

Descriptive

Summarizing what you have

  • Central tendency: mean, median, mode
  • Spread: range, variance, standard deviation, IQR
  • Shape: skewness & kurtosis of the distribution
  • Position: percentiles, z-scores
Inferential

Reasoning beyond the sample

  • Sampling distribution: how a statistic varies across many samples
  • Standard error: the SD of that sampling distribution
  • Confidence interval: a range of plausible parameter values
  • p-value: probability of data this extreme if H₀ is true
What a p-value is not. It is not the probability that H₀ is true, nor the probability the result was due to chance, nor a measure of effect size. It only quantifies how surprising the data are under the null model. A tiny p can accompany a trivial effect in a large sample.

Statistical significance vs. practical significance

Significance tells you an effect is unlikely to be zero. Effect size tells you whether it's big enough to matter. Report both. Drag the sample size below to see why.

Same effect size (Cohen's d = 0.20, a small effect) — only the sample changes.
Sample size (n per group)
30
Approx. p-value
Significant at α = .05?
Effect size (Cohen's d)
0.20

Cohen's d

Standardized mean difference

Difference between two means in SD units.
Benchmarks: 0.2 small · 0.5 medium · 0.8 large.

Correlation r / r²

Strength of association

r ranges −1 to +1; r² is variance explained.
Benchmarks: .10 small · .30 medium · .50 large.

Eta-squared η²

Effect size for ANOVA

Proportion of variance in the DV explained by the grouping factor.
Benchmarks: .01 · .06 · .14.

Modeling

Regression, in depth

Regression fits a line (or hyperplane) that predicts a continuous outcome from one or more predictors. It is the workhorse of prediction and of statistical control — but its output only means what it claims when its assumptions hold.

Simple

One predictor

Ŷ = b₀ + b₁X. The model has a single slope. b₁ is how much Ŷ changes for a one-unit rise in X; b₀ (intercept) is the predicted Ŷ when X = 0.

Equivalent to a Pearson correlation re-expressed in the outcome's raw units.

Multiple

Several predictors

Ŷ = b₀ + b₁X₁ + b₂X₂ + … Each slope is a partial coefficient: the effect of that predictor holding the others constant. This is how regression "controls for" covariates statistically.

Standardized β coefficients put predictors on a common scale for comparison.

Interpreting coefficients & R²

  • Unstandardized b: change in Y per one-unit change in X, in raw units.
  • Standardized β: the same in standard-deviation units — comparable across predictors.
  • R²: the proportion of variance in Y explained by the model (0–1). R² = .30 means the predictors account for 30% of the outcome's variance.
  • Adjusted R²: penalizes R² for extra predictors, so it doesn't reward throwing in junk variables.
  • Statistical significance of a coefficient (its p-value) is separate from its size — report both, ideally with confidence intervals.
Watch multicollinearity. Highly correlated predictors inflate standard errors and make individual slopes unstable and hard to interpret (check the VIF).

The key OLS assumptions (LINE)

AssumptionMeaningHow to check
LinearityThe relationship between predictors and outcome is linearResiduals-vs-fitted plot shows no curve
IndependenceResiduals are not correlated with each otherDesign (no clustering); Durbin–Watson
HomoscedasticityResidual variance is constant across fitted valuesResidual plot shows even band, not a fan
Normality of residualsErrors are approximately normalQ–Q plot / histogram of residuals

When the outcome is binary, these break down and logistic regression (modeling log-odds) replaces OLS.

Instrumentation

Survey & measurement design

A survey is a measurement instrument, and bad items produce numbers that are precise but wrong. Good design starts before a single response is collected — with items that are clear, neutral, and mapped to the construct.

Writing good items

  • One idea per item. Avoid double-barreled questions ("Is the course engaging and useful?") — a respondent can't answer both at once.
  • Neutral wording. Avoid leading and loaded language that pushes toward an answer.
  • Plain language. No jargon, double negatives, or vague quantifiers ("often," "usually").
  • Answerable. Respondents must actually know and be willing to report the answer.
  • Exhaustive, mutually exclusive options for closed items; offer "N/A" or "prefer not to say" where honest.

Response scales

  • Likert scale: symmetric agree–disagree options (often 5 or 7 points) with a neutral midpoint; treated as ordinal, sometimes as interval when summed across items.
  • Balance the anchors: equal positive and negative options, clearly labeled.
  • Odd vs. even points: an odd number allows a true neutral; an even number (forced choice) removes fence-sitting.
  • Consistent direction, but consider a few reverse-scored items to catch inattentive straight-lining.

Common sources of bias to design out

BiasWhat happensGuard against it
Social desirabilityRespondents answer to look good rather than truthfullyAssure anonymity; word sensitive items neutrally
AcquiescenceTendency to agree regardless of contentMix item direction; avoid agree/disagree overuse
Leading / loaded wordingItem nudges toward a particular responsePilot test; use neutral, balanced phrasing
Order & context effectsEarlier items shape answers to later onesRandomize item order where feasible
Nonresponse biasThose who don't respond differ systematicallyMaximize response rate; compare respondents to frame
Central tendency / straight-liningRespondents cluster at the midpoint or repeat one columnReverse-scored items; attention checks
Pilot before you deploy. Cognitive interviewing and a small pilot surface confusing items, floor/ceiling effects, and weak reliability while they are still cheap to fix (Fowler, 2014).
Interactive

Choose the right test

Answer three questions about your analytic goal. This suggests a common starting test — a heuristic, not a substitute for checking each test's assumptions.

1 · What is your analytic goal?

2 · How many groups or predictors?

3 · What kind of outcome variable?

All three tests below assume their conditions are met (e.g. approximate normality, independence, expected cell counts). When assumptions fail, a nonparametric counterpart — Mann–Whitney, Kruskal–Wallis, Spearman — is the defensible substitute.

Reference

Common tests at a glance

A quick map from question to test. Match your goal, number of groups, and outcome type to the row that fits.

TestWhen to useExample questionEffect size
Independent t-testCompare the means of two independent groups on a continuous outcomeDo students in the tutoring program score higher than those who are not?Cohen's d
Paired t-testCompare two related means (pre/post, matched pairs)Did scores rise from pretest to posttest for the same students?Cohen's d
One-way ANOVACompare means across three or more groupsDo three teaching methods differ in average achievement?Eta-squared η²
Pearson correlationMeasure the linear association between two continuous variablesIs study time related to exam performance?r / r²
Linear regressionPredict a continuous outcome from one or more predictorsHow well do attendance and prior GPA predict final grade?R², β
Chi-squareTest association between categorical variables / goodness of fitIs program completion related to enrollment status?Cramér's V, φ
Logistic regressionPredict a binary categorical outcomeWhat predicts whether a student is retained vs. promoted?Odds ratio
Factorial / two-way ANOVATest two or more factors and their interaction on a continuous outcomeDo method and grade level interact to affect achievement?Partial η²
Repeated-measures ANOVACompare three or more related means (same subjects over time)Do scores change across three testing occasions?Partial η²
ANCOVACompare group means while statistically controlling a covariateDo methods differ once we adjust for prior GPA?Partial η²
Mann–Whitney UNonparametric alternative to the independent t-test (ordinal / non-normal)Do two groups differ in ranked satisfaction?Rank-biserial r
Wilcoxon signed-rankNonparametric alternative to the paired t-testDid ranked scores shift from pre to post?Matched-pairs r
Kruskal–WallisNonparametric alternative to one-way ANOVA (3+ groups)Do three schools differ in ranked engagement?Epsilon-squared
Spearman's rhoMonotonic association between two ordinal / non-normal variablesIs class rank related to attendance rank?ρ (rho)
Critical appraisal

Reading results critically

A p-value below .05 is not a certificate of truth. The last decade of methodological reckoning — the "replication crisis" — has shown how ordinary analytic choices can manufacture false positives. Reading results well means knowing where the numbers can mislead.

How false positives get manufactured

  • p-hacking: trying analyses, dropping cases, or adding covariates until p dips below .05 — then reporting only what "worked."
  • Multiple comparisons: run 20 independent tests at α = .05 and, by chance alone, roughly one will be "significant." Correct with Bonferroni or FDR control.
  • HARKing: Hypothesizing After the Results are Known — dressing an exploratory finding up as an a priori prediction.
  • Optional stopping: peeking at the data and stopping collection the moment p < .05 inflates the error rate.
  • Publication bias: significant results get published; null results sit in the file drawer, biasing the literature.

What defensible reporting looks like

  • Pre-registration: lodge the hypotheses and analysis plan before seeing data, separating confirmatory from exploratory claims (a Registered Report goes further, peer-reviewing the plan itself).
  • Effect sizes + confidence intervals, not just p-values — the "New Statistics" shifts the question from "is there an effect?" to "how big, and how precisely estimated?" (Cumming, 2014).
  • The ASA statement: a p-value alone does not measure the size or importance of an effect, nor the probability the hypothesis is true (Wasserstein & Lazar, 2016).
  • Report everything: all measures, conditions, exclusions, and analyses — significant or not.
  • Replication: a single study is a data point, not a verdict.
The garden of forking paths. Even without conscious cheating, the many defensible choices in cleaning and analyzing data give researchers enough "degrees of freedom" to find something significant. Pre-specifying the path is the cure.

Spot the flaw — test your critical reading

Each claim below hides a common methodological flaw. Pick the one that best names it.

1 · Observational study: "Students who eat breakfast score higher on state tests, so schools should serve breakfast to raise scores."

2 · "We tested 20 outcomes and found one significant at p < .05, which we report as our key result."

3 · "With 5,000 students, a correlation of r = .04 between an app's use and grades was significant (p < .01), so we recommend the app."

4 · "The treatment–control difference was not significant (p = .20), proving the program has no effect."

Flaws correctly identified
Answer the four items above
Rigor

Reliability, validity, and defensible measurement

A quantitative claim is only as trustworthy as its measures and its transparency. Check the practices your study uses to gauge how defensible it is.

Reliability

Consistency of a measure

Does the instrument yield stable, repeatable scores? Via internal consistency (Cronbach's α), test–retest, inter-rater agreement.

Validity

Does it measure the construct?

Content, construct, criterion (concurrent & predictive) validity. A measure can be reliable yet invalid.

Replication

Does it hold up again?

Independent replication is the ultimate test of a finding — the antidote to false positives and the "replication crisis."

Transparency

Can others check you?

Pre-registration, open data and code, and reporting all analyses combat p-hacking and HARKing.

Rigor self-check — select every practice your study uses

Validated instrumentsMeasures with documented reliability and validity evidence, not ad-hoc items.
Reported reliabilityCronbach's α or equivalent reported for every multi-item scale.
A priori power analysisSample size justified to detect the expected effect before data collection.
Effect sizes + CIs reportedNot just p-values — the magnitude and precision of every effect.
Pre-registrationHypotheses and analysis plan registered before seeing the data.
Assumptions checkedNormality, independence, homogeneity of variance verified before testing.
IRB approval & consentEthics review, informed consent, confidentiality, and data protection.
Open data & materialsData, code, and instruments shared so others can reproduce the analysis.
Estimated rigor coverage
Select practices to score your design
Ethics is methodological. IRB review, voluntary informed consent, confidentiality, and secure data handling are not paperwork — they are conditions for a study to be both defensible and just. Analytic transparency (reporting all outcomes, not just significant ones) is the ethical core of honest inference.
Reference

Core vocabulary

  • Operationalization — defining a construct by the concrete operations used to measure it.
  • Standard error — the standard deviation of a statistic's sampling distribution.
  • Confidence interval — a range of plausible values for a parameter at a stated confidence level.
  • Type I error (α) — rejecting a true null; a false positive.
  • Type II error (β) — failing to reject a false null; a false negative.
  • Statistical power — the probability of detecting a true effect (1 − β).
  • Effect size — the standardized magnitude of a result, independent of n.
  • p-value — probability of data this extreme if the null were true.
  • Confounding variable — a third variable distorting an observed IV–DV relationship.
  • Generalizability — the extent findings extend beyond the studied sample.

Foundational scholars & further reading

Field — Discovering Statistics Using IBM SPSS Statistics Creswell & Creswell — Research Design Creswell — Educational Research Cohen — Statistical Power Analysis for the Behavioral Sciences Shadish, Cook & Campbell — Experimental & Quasi-Experimental Designs Salkind — Statistics for People Who (Think They) Hate Statistics Gravetter & Wallnau — Statistics for the Behavioral Sciences Tabachnick & Fidell — Using Multivariate Statistics Cohen (1988) — Statistical Power Analysis for the Behavioral Sciences Cumming (2014) — The New Statistics: Why and How Wasserstein & Lazar (2016) — ASA Statement on p-Values Nunnally & Bernstein (1994) — Psychometric Theory Fowler (2014) — Survey Research Methods
▶ View Source