This is the whole vocabulary Statistitch matches on. Every analysis carries two definitions — one for the methodologist, one for everyone else — and a stable identity, so it can be weighted, filtered, and renamed as a field's language shifts without stranding a single profile.
Plain
Gives you a basic summary of one variable at a time—like finding the average score or seeing how much people's responses spread out.
Technical
Calculates central tendency (means, medians, and modes) and variability (variances, skewness, and kurtosis) to summarize single-variable distributions. Helps researchers establish baseline data properties and identify potential anomalies early in the analysis process.
Plain
Checks if two things are related to each other—like whether study hours go up when test scores go up.
Technical
Evaluates the strength and direction of associations between two continuous or categorical variables using metrics like Pearson correlations, Spearman correlation, or Chi-Square. Essential for preliminary hypothesis checking and uncovering raw pairwise relationships.
Plain
Measures how often a health condition occurs in a group of people and who is most at risk.
Technical
Quantifies disease frequency and occurrence across populations using measures like incidence, prevalence, risk ratios and attributable risk. Provides essential foundational metrics for public health surveillance and observational research.
Plain
Turns your numbers into charts and pictures so you can spot trends, odd values, or missing info at a glance.
Technical
Uses graphical representations to profile data distributions, inspect missingness patterns, and detect multivariate outliers visually. Enhances data understanding before formal inferential modeling is conducted.
Plain
A health check for your data to make sure it is clean and properly formatted before running advanced tests.
Technical
Assesses statistical assumptions such as univariate normality, multicollinearity, and homoscedasticity, prior to modeling. Ensures that downstream inferential tests produce valid unbiased parameter estimates.
Plain
Fills in the blanks when you have incomplete data using advanced predictions, so you don't have to throw out valuable records.
Technical
Employs Multiple Imputation (MI) or Full Information Maximum Likelihood (FIML) to handle Missing at Random (MAR) data.
Plain
Tests whether two or more groups have noticeably different average scores (like comparing test results of Group A vs. Group B).
Technical
Compares mean differences across two or more groups using t-tests, ANOVA, or ANCOVA (controlling for continuous covariates). Ideal for experimental and quasi-experimental designs meeting distributional assumptions.
Plain
Compares groups when your data is messy, ranked, or doesn't fit a standard bell curve.
Technical
Tests group differences when data violate normality or ordinal scaling using methods like Mann-Whitney U or Kruskal-Wallis. Provides robust inferences without relying on strict parametric distributional assumptions.
Plain
Compares how the exact same group of people changes over time—like tracking patient progress before during and after treatment.
Technical
Analyzes changes across multiple time points or within-subject conditions using Repeated Measures ANOVA or MANOVA. Accounts for correlated errors inherent in tracking the same subjects repeatedly.
Plain
Checks for connections between two yes/no questions—especially when working with small medical or lab sample sizes.
Technical
Examines categorical relationships in contingency tables using specialized exact methods like Fisher's Exact or McNemar's tests. Crucial for sparse biological data, matched case-control studies, and small clinical samples.
Plain
Adjusts your comparisons so you don't accidentally claim a false discovery when running a lot of tests at once.
Technical
Adjusts significance thresholds using techniques such as Tukey HSD, Bonferroni, or False Discovery Rate (FDR) control. Prevents Inflation of Type I error rates when performing multiple statistical tests simultaneously.
Plain
Predicts a specific number (like salary or test scores) based on other characteristics (e.g., characteristics of a person).
Technical
Estimates linear relationships between continuous outcomes and one or more predictor variables. Widely used for predicting outcomes, hypothesis testing, and hierarchical model comparison.
Plain
Predicts yes/no outcomes from other characteristics.
Technical
Extends linear regression to binary responses using logistic or probit models. Allows flexible modeling of non-linear link functions and overdispersed data.
Plain
Predicts outcomes with multiple categories—like survey ratings (low/medium/high) or choice selection (Option A B or C).
Technical
Models categorical dependent variables with more than two levels whether ordered or unordered. Essential for survey responses preference modeling and multi-category disease staging.
Plain
Predicts yes/no outcomes or counts (like whether someone clicks an ad or how many doctor visits they make).
Technical
Extends linear regression to non-normal response variables including binary (logistic or probit) and Poisson count models. Allows flexible modeling of non-linear link functions and overdispersed data.
Plain
Explains how or why a cause leads to an effect (like how studying leads to better grades because it increases knowledge), or figures out when that effect is strongest (like if studying helps more for math exams than for history exams).
Technical
Evaluates whether a third variable transmits the effect of an independent variable on a dependent variable (mediation) or whether it changes the strength and direction of that relationship (moderation) using path analytic frameworks or interaction terms.
Plain
Analyzes data that is grouped or at different levels-like tracking students within different classrooms, or patients within different hospitals. It lets you understand trends for the whole group while adjusting for the fact that each sub-group might have its own unique starting point or trajectory.
Technical
Accounts for nested or hierarchical data structures by incorporating both fixed effects and random effects. It estimates variance components at multiple levels and handles the non-independence of observations inherent in clustered data, allowing for random intercepts and random slopes across groups.
Plain
Predicts the specific number of times something will happen—like how many visits a customer will make to a store.
Technical
Models count-based outcome variables using Poisson or Negative Binomial regression to account for overdispersion and zero-inflation.
Plain
Forecasts future trends based on past historical patterns—like predicting seasonal sales or daily traffic.
Technical
Models time-dependent processes to account for autocorrelation, seasonal variations, and trends using ARIMA and SARIMA frameworks.
Plain
Cleans up jumpy data over time to reveal the clear underlying direction and seasonal patterns.
Technical
Breaks down sequential data into underlying trend seasonal and irregular noise components using techniques like Holt-Winters or STL decomposition.
Plain
Looks at how multiple changing timelines influence each other over time (like inflation and interest rates).
Technical
Analyzes dynamic interrelationships among multiple interdependent time series using Vector Autoregressions (VAR) and Granger causality.
Plain
Measures whether a specific event or new law caused a sudden shift in ongoing trends.
Technical
Evaluates the effect of discrete policy changes interventions or events over time using Interrupted Time Series Analysis (ITSA). Controls for pre-existing trends to estimate true post-intervention impact.
Plain
Tracks hidden underlying changes in real-time continuous data (like monitoring sensor feeds or vital signs).
Technical
Tracks underlying hidden states that evolve dynamically over time using algorithms like Kalman filtering. Highly effective for noisy continuous monitoring data and tracking non-stationary processes.
Plain
Tracks how long it takes for a specific event to happen—like how long a medicine keeps a illness in remission.
Technical
Estimates survival probabilities over time without parametric assumptions using Kaplan-Meier curves and Log-Rank tests. These are standard approaches for visualizing time-to-event trajectories across study cohorts.
Plain
Calculates how much different risk factors increase or decrease the chances of an event happening over time.
Technical
Models hazard rates while controlling for covariates using Cox Proportional Hazards or parametric acceleration models. Quantifies risk ratios for clinical outcomes adjusted for participant risk profiles.
Plain
Sets up fair clinical testing groups so that treatment results aren't biased by who ended up in which group.
Technical
Establishes experimental protocols including parallel crossover and factorial trial designs with stratified block randomization. Minimizes selection bias and ensures internal validity in clinical interventions.
Plain
Proves that a new cheaper or safer treatment works just as well as the existing standard treatment.
Technical
Determines whether a new treatment is not unacceptably worse than standard care using Two One-Sided Tests (TOST). Critical for bioequivalence studies and pharmaceutical drug approval processes.
Plain
Lets researchers check medical trial data early to stop the study if a drug is clearly working or unsafe.
Technical
Evaluates accumulating trial data at pre-planned intervals using group sequential stopping boundaries like O'Brien-Fleming. Allows trials to stop early for efficacy or safety without inflating Type I error rates.
Plain
Measures how accurately a medical test gives correct positive and negative results.
Technical
Evaluates diagnostic performance metrics including ROC curve analysis sensitivity specificity and AUC comparisons. Quantifies how effectively diagnostic tools or biological markers classify disease states.
Plain
Designs and tests questionnaires to make sure the questions are clear, unbiased, and actually measure what you want them to measure without confusing the people taking the survey.
Technical
Constructs and validates self-report measurement instruments using cognitive interviewing, pilot testing, and initial psychometric evaluation to ensure content validity, construct validity, and reliability of latent trait measurement prior to full-scale deployment.
Plain
Checks whether a group of survey questions actually measures the single big concept you intended (like anxiety or satisfaction).
Technical
Explores or confirms latent construct structures underlying observed indicator items via EFA and CFA. Essential for scale development instrument validation and assessing construct validity.
Plain
Measures how consistent and reliable a test or questionnaire is overall.
Technical
Evaluates score reliability and item characteristics using metrics such as Cronbach's alpha, McDonald's omega, and item-total correlations. Standard methodology for assessing internal consistency in educational and psychological measurement.
Plain
Analyzes individual test questions to see how hard they are and whether they treat all student groups fairly.
Technical
Models item-level properties and individual latent trait levels using 1PL, 2PL, and 3PL models. Used for adaptive testing test assembly and ensuring item fairness across groups.
Plain
Ensures a survey or test means the exact same thing to different groups of people (like across different languages or ages).
Technical
Tests whether psychological or educational constructs maintain consistent measurement properties across diverse demographic groups or time points. Prevents construct bias when comparing group means.
Plain
Determines the official passing scores or grade boundaries for a test. It is the structured, research-backed process used to decide exactly how many points someone needs to earn to pass an exam or be classified at a specific skill level (like "basic," "proficient," or "advanced").
Technical
Establishes psychometrically defensible cut scores for criterion-referenced assessments using formalized, evidence-based methodologies such as the Modified Angoff, Bookmark, or Contrasting Groups methods. Ensures that categorical classifications (e.g., pass/fail, proficiency levels) are reliable, valid, and legally robust for high-stakes decision making.
Plain
Creates smart tests that adjust the difficulty of questions on the fly based on how well the test-taker is doing, making the test shorter and more accurate.
Technical
Utilizes IRT algorithms to dynamically tailor item selection to the examinee's estimated latent trait level in real-time, optimizing measurement precision.
Plain
Instead of just giving an overall test score (like an 85%), this method looks at exactly how someone answered the questions to map out their specific strengths and weaknesses. It breaks down performance to pinpoint exactly which specific skills or concepts a person has mastered and which ones they still need help with—much like a doctor diagnosing specific symptoms rather than just telling you that you are sick.
Technical
A class of multidimensional, discrete latent variable models used to classify individuals into categorical latent profiles based on their item responses. Utilizing a Q-matrix that maps specific items to fine-grained underlying attributes, CDMs/DCMs estimate the probability of an examinee possessing specific combinations of unobservable skills or traits, providing detailed diagnostic feedback across multiple dimensions rather than a single continuous continuum.
Plain
Tests big step-by-step chain reactions where multiple cause-and-effect relationships happen all at once.
Technical
Evaluates complex networks of direct, indirect, and latent relationships simultaneously using path analysis and full structural models. Allows rigorous testing of theoretical mechanisms and mediation hypotheses.
Plain
Finds hidden sub-groups of people in your data who share similar profiles or behaviors.
Technical
Identifies unobserved homogeneous subgroups within heterogeneous populations using LCA, LPA, and Growth Mixture Modeling. Useful for discovering person-centered profiles and distinct development trajectories.
Plain
Matches similar people in real-world data to create fair comparison groups when a randomized trial isn't possible.
Technical
Reduces selection bias in observational studies by balancing baseline covariates across treatment groups via matching, IPTW, or stratification. Simulates experimental conditions when random assignment is impossible.
Plain
Uses clever real-world setups to figure out if a policy or program actually caused a change.
Technical
Estimates causal effects using natural experiments through Difference-in-Differences, Regression Discontinuity, or Instrumental Variables. Provides credible causal estimates when randomization is unethical or impractical.
Plain
Calculates exactly how many participants you need to recruit so your study produces trustworthy results.
Technical
Determines required sample sizes to detect meaningful effect sizes using power calculations and Monte Carlo simulations. Ensures study designs are adequately powered to minimize Type II errors.
Plain
Teaches a computer to recognize patterns from past data so it can automatically classify new examples.
Technical
Trains predictive algorithms like Random Forests, XGBoost, and Support Vector Machines on labeled data. Produces high-accuracy predictions for complex non-linear outcomes.
Plain
Lets a computer group similar data points together on its own without telling it what to look for beforehand.
Technical
Groups unlabeled data points or reduces dimensionality using k-Means clustering, PCA, t-SNE, or UMAP. Reveals hidden structure patterns and clusters in complex multidimensional datasets.
Plain
Sorts through massive datasets with thousands of measurements per person (like gene data) to find the key signals.
Technical
Processes biological datasets with thousands of parallel features using differential expression analysis and regularized regression (Lasso/Ridge/Elastic Net). Controls false discovery rates in high-throughput genomic or proteomic studies.
Plain
Re-tests your data hundreds of times in different ways to make sure your conclusions are solid and reliable.
Technical
Derives empirically calibrated standard errors and confidence intervals using bootstrapping cross-validation and robust estimators. Safeguards statistical inferences when distributional assumptions are violated.
Plain
Combines results from many different studies on the same topic to find the overall consensus and prove whether a finding is truly reliable.
Technical
Synthesizes effect sizes across multiple independent studies using fixed-effects or random-effects models, testing for heterogeneity and publication bias.
Plain
Calculates the probability of an event happening by combining what we already know with new evidence as it comes in.
Technical
Updates prior probability distributions with observed data to generate posterior distributions using methods that include Markov Chain Monte Carlo (MCMC) estimation and Bayesian approaches.
Plain
Scans thousands of written comments or reviews to automatically find common themes or figure out if people are generally happy or upset.
Technical
Applies computational linguistics to extract topics, sentiment, and latent semantic structures from unstructured text data.
Plain
Looks at data on a map to see if things happen in geographic clusters—like tracking how a disease spreads through different neighborhoods.
Technical
Evaluates geographic patterns and spatial dependency using Moran's I, spatial autoregressive models, and geospatial mapping.