Statistics
Statistics glossary
Terms for describing data, drawing inferences, building models and reasoning about cause, with the common ways statistics mislead.
B
- Base rate neglect also: base rate fallacy
The tendency to ignore the prior frequency of an event when judging the probability of a case, relying instead on specific information such as a test result. It leads people to overestimate the chance that a positive test means disease when the disease is rare. Presenting information as natural frequencies (for example, 9 out of 1,000) greatly reduces the error. (Bar-Hillel, 1980; Gigerenzer & Hoffrage, 1995)
- Bayes' theorem
A rule for updating the probability of a hypothesis in light of evidence: the posterior probability is proportional to the prior probability times the likelihood of the evidence under the hypothesis. Formally, P(H | E) = P(E | H) × P(H) / P(E). It underlies Bayesian statistics and much of rational belief revision. (Bayes, 1763; Jaynes, 2003)
- Bayesian inference
An approach to statistics in which unknown quantities are treated as having probability distributions. A prior distribution expressing what is known before the data is combined with the likelihood of the data to give a posterior distribution, from which estimates and credible intervals are read directly as probability statements about the parameters. (Edwards et al., 1963; Jaynes, 2003)
- Bootstrap
A resampling method introduced by Bradley Efron in 1979 that estimates the sampling distribution of a statistic by repeatedly drawing samples, with replacement, from the observed data and recomputing the statistic each time. The spread of the resampled values gives standard errors and confidence intervals without strong distributional assumptions. (Efron, 1979)
C
- Central limit theorem
The result that the distribution of the mean (or sum) of many independent observations with finite variance approaches a normal distribution as the sample size grows, whatever the shape of the original distribution. It explains why the normal distribution is used so widely for sampling distributions and confidence intervals. (Lohr, 2021)
- Chi-square test
A family of tests introduced by Karl Pearson in 1900 that compare observed frequencies with those expected under a hypothesis. The test of independence asks whether two categorical variables in a cross-tabulation are associated. With complex survey samples, design-adjusted versions such as the Rao–Scott correction are needed. (Pearson, 1900)
- Cluster analysis
Methods that group cases so that cases in the same cluster are more similar to each other than to those in other clusters, using distances such as Euclidean distance. K-means and hierarchical clustering are common. In market research it is used to build segments, whose usefulness depends on the variables chosen. (Hastie et al., 2009)
- Cohen's kappa
A measure of agreement between two raters assigning cases to categories, corrected for the agreement expected by chance. Kappa is 1 for perfect agreement and 0 for chance-level agreement. It is widely used to check the reliability of coding in content analysis and qualitative research. (Cohen, 1960)
- Collider bias
Bias that arises when an analysis conditions on (selects on, stratifies by or adjusts for) a variable that is caused by both the exposure and the outcome. Conditioning on such a collider creates a spurious association between its causes, as in Berkson's finding that two diseases appear related in hospital samples. (Elwert & Winship, 2014; Berkson, 1946; Hernán et al., 2004)
- Confidence interval
A range computed from sample data by a procedure that, over repeated samples, contains the true parameter a stated proportion of the time, such as 95%. A particular interval either contains the parameter or does not; the 95% describes the procedure, not the probability for one interval. Jerzy Neyman set out the theory in 1937. (Neyman, 1937; Greenland et al., 2016)
- Confounding
Distortion of the estimated effect of an exposure on an outcome because a third variable causes both. Confounding can create, hide or reverse an association. It is controlled by randomisation, or in observational data by adjustment, matching or stratification on the confounders, which requires that they are known and measured. (Pearl, 2009; Hernán et al., 2004)
- Correlation coefficient
A number between −1 and +1 summarising the strength and direction of association between two variables. Pearson's r measures linear association; Spearman's rho measures monotonic association using ranks. A correlation does not show causation and can be distorted by outliers and by non-linear relationships, as Anscombe's quartet illustrates. (Anscombe, 1973)
- Counterfactual
The outcome that would have occurred under a condition other than the one that actually happened. In the potential outcomes framework each unit has an outcome under treatment and under control, and the causal effect is their difference. Because only one is ever observed, causal inference is a problem of estimating the missing counterfactual. (Rubin, 1974; Holland, 1986)
- Cronbach's alpha
A coefficient of internal consistency for a multi-item scale, based on the number of items and the average inter-item covariance. Values closer to 1 indicate that items vary together. A high alpha does not prove that a scale measures a single construct, and alpha rises with the number of items. (Cronbach, 1951)
- Cross-validation
A method for estimating how well a model will predict new data by repeatedly fitting it on part of the data and testing it on the held-out remainder. In k-fold cross-validation the data are split into k parts and each part serves once as the test set. It is used to choose model complexity and avoid overfitting. (Hastie et al., 2009)
D
- Difference-in-differences also: DiD
A method that estimates a causal effect by comparing the change in an outcome over time in a group exposed to a policy or event with the change in an unexposed group. It removes stable differences between the groups and common trends, under the key assumption that the groups would have followed parallel trends without the intervention. (Card & Krueger, 1994; Angrist & Pischke, 2009)
- Differential privacy
A mathematical definition of privacy for statistical releases: an algorithm is differentially private if its output distribution changes very little whether or not any one person's data are included. It is usually achieved by adding calibrated random noise, and it gives a measurable privacy budget, epsilon. The US Census Bureau used it for 2020 census products. (Dwork et al., 2006)
- Directed acyclic graph also: DAG
A diagram of variables (nodes) connected by arrows representing direct causal effects, with no path that returns to its starting point. DAGs make causal assumptions explicit and show which variables must be adjusted for to estimate an effect (the back-door criterion) and which must not, such as colliders and mediators. (Pearl, 2009; Pearl & Mackenzie, 2018)
E
- Ecological fallacy
Inferring relationships about individuals from data on groups. William Robinson showed in 1950 that the correlation between literacy and foreign birth across US states had the opposite sign to the correlation among individuals. Group-level correlations can be strong even when individual-level ones are weak or reversed. (Robinson, 1950)
- Effect size
A quantitative measure of the magnitude of a difference or relationship, such as a mean difference, Cohen's d, an odds ratio or a correlation. Unlike a p-value it does not depend on sample size, so it indicates practical importance. Jacob Cohen's conventions of small, medium and large effects are widely used but were meant only as a fallback. (Cohen, 1988)
F
- Factor analysis
A family of methods that explain the correlations among many observed variables by a smaller number of unobserved (latent) factors. Exploratory factor analysis looks for structure; confirmatory factor analysis tests a structure specified in advance. Charles Spearman introduced the approach in 1904 in his work on general intelligence. (Spearman, 1904)
- False discovery rate also: FDR
The expected proportion of rejected null hypotheses that are actually true, that is, the share of "discoveries" that are false. The Benjamini–Hochberg procedure controls it when many tests are performed, offering more power than methods that control the chance of any false positive. (Benjamini & Hochberg, 1995)
- Frequentist inference
An approach to statistics in which probability means long-run relative frequency, parameters are fixed but unknown, and methods are judged by how they would perform over repeated samples. Confidence intervals, p-values and hypothesis tests are frequentist tools. (Neyman, 1937)
G
- Generalised linear model also: GLM
An extension of linear regression, introduced by Nelder and Wedderburn in 1972, in which the outcome can follow any distribution from the exponential family (normal, binomial, Poisson and others) and is related to the predictors through a link function. Logistic and Poisson regression are special cases. (Nelder & Wedderburn, 1972)
I
- Instrumental variable also: IV
A variable that affects the treatment, is unrelated to unmeasured confounders, and affects the outcome only through the treatment. Under these assumptions it can be used to estimate a causal effect from observational data, typically the effect for "compliers" whose treatment is changed by the instrument. (Angrist et al., 1996)
- Inter-rater reliability
The degree to which different coders or raters give consistent judgements of the same material. Simple percentage agreement overstates it because some agreement occurs by chance; chance-corrected statistics such as Cohen's kappa or Krippendorff's alpha are preferred. (Cohen, 1960; Krippendorff, 2019)
L
- Lasso
Least absolute shrinkage and selection operator: a regression method that adds a penalty proportional to the sum of the absolute values of the coefficients. The penalty shrinks coefficients towards zero and sets some exactly to zero, so it performs variable selection and reduces overfitting when there are many predictors. (Tibshirani, 1996; Hastie et al., 2009)
- Latent class analysis
A statistical model that assumes the population consists of a number of unobserved classes and estimates, from categorical survey answers, the probability of each answer in each class and each respondent's probability of class membership. It is a model-based alternative to cluster analysis for segmentation. (Hastie et al., 2009)
- Law of large numbers
The theorem that the average of independent observations from the same distribution converges to the expected value as the number of observations grows. It explains why large random samples give stable estimates, and it does not imply that short runs "even out", the gambler's fallacy. (Jaynes, 2003)
- Levels of measurement also: scales of measurement
Stanley Stevens's classification of variables as nominal (categories), ordinal (ordered categories), interval (equal differences, arbitrary zero) or ratio (equal differences, true zero). The level determines which statistics are meaningful, although applying the rules strictly, for example to rating scales, remains debated. (Stevens, 1946)
- Logistic regression
A regression model for a binary outcome in which the log-odds of the outcome is a linear function of the predictors. Exponentiated coefficients are odds ratios. It is a generalised linear model with a binomial distribution and a logit link. (Nelder & Wedderburn, 1972)
M
- Model selection criteria also: AIC, BIC
Measures for choosing among statistical models that balance fit against complexity. The Akaike information criterion (AIC) estimates out-of-sample prediction error; the Bayesian information criterion (BIC) penalises extra parameters more heavily as sample size grows. Lower values indicate a preferred model. (Akaike, 1974)
- Monte Carlo simulation
Using repeated random sampling to approximate quantities that are hard to calculate exactly, such as the distribution of outcomes of a risky project or the behaviour of a statistical method. Metropolis and Ulam named and described the method in 1949, drawing on work at Los Alamos. (Metropolis & Ulam, 1949)
- MRP also: multilevel regression and post-stratification
Multilevel regression and post-stratification: a method that fits a multilevel model predicting a survey outcome from demographic and geographic variables, then weights the predictions for each population cell by its known size from census data. It produces estimates for small groups and areas and can correct for unrepresentative samples. (Park et al., 2004)
- Multilevel model also: hierarchical model
A regression model for data with a nested structure, such as employees within firms or repeated measures within people, that includes random effects at each level. It estimates variation within and between groups and "partially pools" estimates for small groups towards the overall mean. (Gelman & Hill, 2007)
- Multiple comparisons
The problem that when many hypothesis tests are performed, the chance that at least one gives a false positive grows well above the nominal significance level. Corrections such as Bonferroni (controlling the family-wise error rate) and Benjamini–Hochberg (controlling the false discovery rate) adjust for it. (Benjamini & Hochberg, 1995)
N
- Normal distribution also: Gaussian distribution
The symmetric, bell-shaped continuous distribution defined by its mean and standard deviation, with about 68% of values within one standard deviation of the mean and about 95% within two. Its importance comes largely from the central limit theorem. (Lohr, 2021)
O
- Odds ratio
The ratio of the odds of an outcome in one group to the odds in another; a value of 1 means no association. It is the natural effect measure in logistic regression and case–control studies. When the outcome is common, an odds ratio exaggerates the corresponding risk ratio. (Nelder & Wedderburn, 1972)
- Overfitting
Fitting a model so closely to the particular sample that it captures noise as well as signal, so that it performs well on the training data but poorly on new data. It is the high-variance side of the bias–variance trade-off and is controlled by cross-validation, regularisation and simpler models. (Hastie et al., 2009)
P
- Potential outcomes also: Rubin causal model
A framework for causal inference, associated with Donald Rubin, in which each unit has a potential outcome under each possible treatment. The causal effect is the difference between them, and since only one is observed, the "fundamental problem of causal inference", effects are estimated by comparing groups under assumptions such as random assignment. (Rubin, 1974; Holland, 1986)
- Principal component analysis also: PCA
A dimension-reduction method, first described by Karl Pearson in 1901, that transforms correlated variables into a smaller set of uncorrelated components that capture as much of the total variance as possible. Unlike factor analysis it does not model a latent cause of the variables. (Pearson, 1901)
- Prior probability also: prior
In Bayesian reasoning, the probability assigned to a hypothesis before the current evidence is considered, often based on base rates or earlier studies. Combined with the likelihood of the evidence, it gives the posterior probability. Neglecting the prior is the source of base rate neglect. (Bayes, 1763; Jaynes, 2003)
- Propensity score
The probability that a unit receives a treatment given its observed characteristics. Rosenbaum and Rubin showed that matching, weighting or stratifying on this single score can balance the observed covariates between treated and untreated groups. It does not correct for unmeasured confounders. (Rosenbaum & Rubin, 1983)
- P-value
The probability, computed assuming the null hypothesis and all model assumptions are true, of obtaining a result at least as extreme as the one observed. It is not the probability that the null hypothesis is true, nor a measure of effect size. The American Statistical Association's 2016 statement warns against basing conclusions on whether p crosses 0.05. (Wasserstein & Lazar, 2016; Greenland et al., 2016)
R
- Regression
A family of methods that model how the expected value of an outcome changes with one or more predictors. In linear regression, coefficients estimate the change in the outcome per unit change in a predictor, holding the other predictors constant. Galton coined the term from his observation of regression to the mean in heredity. (Galton, 1886; Gelman & Hill, 2007)
- Regression discontinuity design also: RDD
A quasi-experimental design used when treatment is assigned by whether a score crosses a cut-off, such as a test mark for a scholarship. Comparing units just above and just below the threshold estimates the causal effect at the cut-off, because those units are otherwise similar. (Thistlethwaite & Campbell, 1960; Angrist & Pischke, 2009)
- Regression to the mean
The tendency for unusually high or low measurements to be followed by values closer to the average, because extreme values partly reflect chance. Francis Galton described it in heredity. Ignoring it leads people to credit interventions, or blame them, for changes that would have happened anyway. (Galton, 1886; Kahneman & Tversky, 1973)
S
- Sampling distribution
The distribution of a statistic, such as a sample mean, across all possible samples of the same size from a population. Its standard deviation is the standard error. It is the basis of confidence intervals and hypothesis tests. (Lohr, 2021)
- Selection bias
Bias that arises when the units included in an analysis differ systematically from the population of interest because of how they were selected, or because selection depends on both exposure and outcome. Coverage error, nonresponse bias, survivorship bias and collider bias are forms of it. (Hernán et al., 2004)
- Simpson's paradox
A reversal in which an association between two variables in every subgroup disappears or reverses when the subgroups are combined. The Berkeley graduate admissions data are a classic example. Which comparison is correct depends on the causal structure, not on the arithmetic. (Simpson, 1951; Bickel et al., 1975; Pearl, 2009)
- Small area estimation
Methods for producing reliable estimates for geographic areas or groups where the survey sample is too small for direct estimates. Models borrow strength from other areas and from auxiliary data such as census or administrative records. (Rao & Molina, 2015)
- Spurious correlation
An association between two variables that does not reflect a causal link between them, because both are driven by a third factor, because of chance, because of shared trends over time, or because of how the variables were constructed. Karl Pearson used the term in 1897 for correlations between ratios with a common denominator. (Pearson, 1897; Yule, 1926; Messerli, 2012)
- Standard deviation
A measure of spread equal to the square root of the variance, the average squared distance of values from their mean. It is expressed in the units of the data. For roughly normal data, about two-thirds of values lie within one standard deviation of the mean. (Tukey, 1977)
- Standard error
The standard deviation of the sampling distribution of a statistic, measuring how much the statistic would vary from sample to sample. For a mean from a simple random sample it equals the standard deviation divided by the square root of the sample size. Complex designs require design-based variance estimation. (Lohr, 2021; Kish, 1965)
- Statistical power
The probability that a test will detect an effect of a given size if it exists, that is, one minus the probability of a Type II error. Power depends on the effect size, the sample size, the significance level and the variability of the data. Underpowered studies miss real effects and exaggerate the effects they do find. (Cohen, 1988)
- Statistical process control also: SPC
Monitoring a process over time with control charts that separate common-cause variation, inherent in a stable process, from special-cause variation that signals a change worth investigating. Walter Shewhart developed the method at Bell Laboratories in the 1920s. (Shewhart, 1931)
- Statistical significance
The label applied when a p-value falls below a chosen threshold, conventionally 0.05. A significant result is not necessarily important, and a non-significant result is not evidence of no effect. Many statisticians recommend reporting effect sizes and confidence intervals instead of dichotomising results. (Wasserstein & Lazar, 2016; Greenland et al., 2016)
- Survivorship bias
Drawing conclusions only from cases that passed a selection process, such as successful firms or returning aircraft, while overlooking those that did not. Abraham Wald's Second World War analysis of where to armour bombers, based on where surviving planes were not hit, is the classic example. (Mangel & Samaniego, 1984)
- Synthetic control method
A method for estimating the effect of an intervention on a single unit, such as a region, by constructing a weighted combination of untreated units that closely matches the treated unit before the intervention. The divergence between the treated unit and its synthetic control afterwards estimates the effect. (Abadie et al., 2010)
T
- t-test also: Student's t-test
A test comparing a mean with a value, or two means with each other, using the t distribution, which accounts for the extra uncertainty from estimating the standard deviation in small samples. William Sealy Gosset published it in 1908 under the pseudonym "Student". (Student, 1908)
- Type I error
Rejecting a null hypothesis that is true, a false positive. Its probability is set by the significance level, often 0.05. A Type II error is failing to reject a false null hypothesis, a false negative; its probability is one minus the power. (Cohen, 1988)
W
- Weighted mean
An average in which each value counts in proportion to a weight, calculated as the sum of each value times its weight divided by the sum of the weights. In surveys the weights represent how many population members each respondent stands for, so the weighted mean estimates the population mean. (Horvitz & Thompson, 1952; Kish, 1992)
References
Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic control methods for comparative case studies: Estimating the effect of California's tobacco control program. Journal of the American Statistical Association, 105(490), 493–505. https://doi.org/10.1198/jasa.2009.ap08746
Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19(6), 716–723. https://doi.org/10.1109/TAC.1974.1100705
Angrist, J. D., & Pischke, J.-S. (2009). Mostly harmless econometrics: An empiricist's companion. Princeton University Press. https://doi.org/10.1515/9781400829828
Angrist, J. D., Imbens, G. W., & Rubin, D. B. (1996). Identification of causal effects using instrumental variables. Journal of the American Statistical Association, 91(434), 444–455. https://doi.org/10.1080/01621459.1996.10476902
Anscombe, F. J. (1973). Graphs in statistical analysis. The American Statistician, 27(1), 17–21. https://doi.org/10.1080/00031305.1973.10478966
Bar-Hillel, M. (1980). The base-rate fallacy in probability judgments. Acta Psychologica, 44(3), 211–233. https://doi.org/10.1016/0001-6918(80)90046-3
Bayes, T. (1763). An essay towards solving a problem in the doctrine of chances. Philosophical Transactions of the Royal Society of London, 53, 370–418. https://doi.org/10.1098/rstl.1763.0053
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57(1), 289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x
Berkson, J. (1946). Limitations of the application of fourfold table analysis to hospital data. Biometrics Bulletin, 2(3), 47–53. https://doi.org/10.2307/3002000
Bickel, P. J., Hammel, E. A., & O'Connell, J. W. (1975). Sex bias in graduate admissions: Data from Berkeley. Science, 187(4175), 398–404. https://doi.org/10.1126/science.187.4175.398
Card, D., & Krueger, A. B. (1994). Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania. American Economic Review, 84(4), 772–793.
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37–46. https://doi.org/10.1177/001316446002000104
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates.
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16(3), 297–334. https://doi.org/10.1007/BF02310555
Dwork, C., McSherry, F., Nissim, K., & Smith, A. (2006). Calibrating noise to sensitivity in private data analysis. In S. Halevi & T. Rabin (Eds.), Theory of cryptography (pp. 265–284). Springer. https://doi.org/10.1007/11681878_14
Edwards, W., Lindman, H., & Savage, L. J. (1963). Bayesian statistical inference for psychological research. Psychological Review, 70(3), 193–242. https://doi.org/10.1037/h0044139
Efron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 1–26. https://doi.org/10.1214/aos/1176344552
Elwert, F., & Winship, C. (2014). Endogenous selection bias: The problem of conditioning on a collider variable. Annual Review of Sociology, 40, 31–53. https://doi.org/10.1146/annurev-soc-071913-043455
Galton, F. (1886). Regression towards mediocrity in hereditary stature. Journal of the Anthropological Institute of Great Britain and Ireland, 15, 246–263. https://doi.org/10.2307/2841583
Gelman, A., & Hill, J. (2007). Data analysis using regression and multilevel/hierarchical models. Cambridge University Press. https://doi.org/10.1017/CBO9780511790942
Gigerenzer, G., & Hoffrage, U. (1995). How to improve Bayesian reasoning without instruction: Frequency formats. Psychological Review, 102(4), 684–704. https://doi.org/10.1037/0033-295X.102.4.684
Greenland, S., Senn, S. J., Rothman, K. J., Carlin, J. B., Poole, C., Goodman, S. N., & Altman, D. G. (2016). Statistical tests, P values, confidence intervals, and power: A guide to misinterpretations. European Journal of Epidemiology, 31(4), 337–350. https://doi.org/10.1007/s10654-016-0149-3
Hastie, T., Tibshirani, R., & Friedman, J. (2009). The elements of statistical learning: Data mining, inference, and prediction (2nd ed.). Springer. https://doi.org/10.1007/978-0-387-84858-7
Hernán, M. A., Hernández-Díaz, S., & Robins, J. M. (2004). A structural approach to selection bias. Epidemiology, 15(5), 615–625. https://doi.org/10.1097/01.ede.0000135174.63482.43
Holland, P. W. (1986). Statistics and causal inference. Journal of the American Statistical Association, 81(396), 945–960. https://doi.org/10.1080/01621459.1986.10478354
Horvitz, D. G., & Thompson, D. J. (1952). A generalization of sampling without replacement from a finite universe. Journal of the American Statistical Association, 47(260), 663–685. https://doi.org/10.1080/01621459.1952.10483446
Jaynes, E. T. (2003). Probability theory: The logic of science (G. L. Bretthorst, Ed.). Cambridge University Press. https://doi.org/10.1017/CBO9780511790423
Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological Review, 80(4), 237–251. https://doi.org/10.1037/h0034747
Kish, L. (1965). Survey sampling. Wiley.
Kish, L. (1992). Weighting for unequal Pi. Journal of Official Statistics, 8(2), 183–200.
Krippendorff, K. (2019). Content analysis: An introduction to its methodology (4th ed.). SAGE.
Lohr, S. L. (2021). Sampling: Design and analysis (3rd ed.). CRC Press. https://doi.org/10.1201/9780429298899
Mangel, M., & Samaniego, F. J. (1984). Abraham Wald's work on aircraft survivability. Journal of the American Statistical Association, 79(386), 259–267. https://doi.org/10.1080/01621459.1984.10478038
Messerli, F. H. (2012). Chocolate consumption, cognitive function, and Nobel laureates. New England Journal of Medicine, 367(16), 1562–1564. https://doi.org/10.1056/NEJMon1211064
Metropolis, N., & Ulam, S. (1949). The Monte Carlo method. Journal of the American Statistical Association, 44(247), 335–341. https://doi.org/10.1080/01621459.1949.10483310
Nelder, J. A., & Wedderburn, R. W. M. (1972). Generalized linear models. Journal of the Royal Statistical Society: Series A (General), 135(3), 370–384. https://doi.org/10.2307/2344614
Neyman, J. (1937). Outline of a theory of statistical estimation based on the classical theory of probability. Philosophical Transactions of the Royal Society of London. Series A, 236(767), 333–380. https://doi.org/10.1098/rsta.1937.0005
Park, D. K., Gelman, A., & Bafumi, J. (2004). Bayesian multilevel estimation with poststratification: State-level estimates from national polls. Political Analysis, 12(4), 375–385. https://doi.org/10.1093/pan/mph024
Pearl, J., & Mackenzie, D. (2018). The book of why: The new science of cause and effect. Basic Books.
Pearl, J. (2009). Causality: Models, reasoning, and inference (2nd ed.). Cambridge University Press. https://doi.org/10.1017/CBO9780511803161
Pearson, K. (1897). Mathematical contributions to the theory of evolution: On a form of spurious correlation which may arise when indices are used in the measurement of organs. Proceedings of the Royal Society of London, 60, 489–498. https://doi.org/10.1098/rspl.1896.0076
Pearson, K. (1900). On the criterion that a given system of deviations from the probable in the case of a correlated system of variables is such that it can be reasonably supposed to have arisen from random sampling. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 50(302), 157–175. https://doi.org/10.1080/14786440009463897
Pearson, K. (1901). On lines and planes of closest fit to systems of points in space. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 2(11), 559–572. https://doi.org/10.1080/14786440109462720
Rao, J. N. K., & Molina, I. (2015). Small area estimation (2nd ed.). Wiley. https://doi.org/10.1002/9781118735855
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals. American Sociological Review, 15(3), 351–357. https://doi.org/10.2307/2087176
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41–55. https://doi.org/10.1093/biomet/70.1.41
Rubin, D. B. (1974). Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology, 66(5), 688–701. https://doi.org/10.1037/h0037350
Shewhart, W. A. (1931). Economic control of quality of manufactured product. Van Nostrand.
Simpson, E. H. (1951). The interpretation of interaction in contingency tables. Journal of the Royal Statistical Society: Series B (Methodological), 13(2), 238–241. https://doi.org/10.1111/j.2517-6161.1951.tb00088.x
Spearman, C. (1904). "General intelligence," objectively determined and measured. The American Journal of Psychology, 15(2), 201–292. https://doi.org/10.2307/1412107
Stevens, S. S. (1946). On the theory of scales of measurement. Science, 103(2684), 677–680. https://doi.org/10.1126/science.103.2684.677
Student. (1908). The probable error of a mean. Biometrika, 6(1), 1–25. https://doi.org/10.1093/biomet/6.1.1
Thistlethwaite, D. L., & Campbell, D. T. (1960). Regression-discontinuity analysis: An alternative to the ex post facto experiment. Journal of Educational Psychology, 51(6), 309–317. https://doi.org/10.1037/h0044319
Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological), 58(1), 267–288. https://doi.org/10.1111/j.2517-6161.1996.tb02080.x
Tukey, J. W. (1977). Exploratory data analysis. Addison-Wesley.
Wasserstein, R. L., & Lazar, N. A. (2016). The ASA statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129–133. https://doi.org/10.1080/00031305.2016.1154108
Yule, G. U. (1926). Why do we sometimes get nonsense-correlations between time-series? A study in sampling and the nature of time-series. Journal of the Royal Statistical Society, 89(1), 1–63. https://doi.org/10.2307/2341482