Småsamplesjustering

Småsamplesjustering, often translated as small-sample adjustment, is a term used in statistics to describe methods and corrections applied when the available data are limited in size. The aim is to improve the reliability of estimates and inferences by mitigating bias, underestimation of variability, and incorrect distributional assumptions that can arise with small samples.

Common applications include estimation of population parameters (mean, proportion), hypothesis testing, and model selection. In small samples, standard approaches that rely on large-sample theory can perform poorly, leading to biased estimates or confidence intervals that are too narrow.

Techniques associated with småsamplesjustering include using the t-distribution with appropriate degrees of freedom or adjustments such as Welch's correction for unequal variances, and exact tests for small contingency tables (Fisher's exact test). Nonparametric or resampling methods (bootstrapping and permutation tests) are also frequently employed. Penalized likelihood methods (for example, Firth’s correction in logistic regression) address estimator bias in small samples. Information criteria for model selection may be adjusted for small samples, such as AICc.

Practically, the choice of method depends on the data structure and model, and small-sample corrections often trade bias against increased variance. Simulation studies and empirical validation are common to assess performance in a given setting.

See also: bias correction, bootstrap, Fisher's exact test, AICc, Welch-Satterthwaite, Firth's correction.