AI-ordlista
Den kompletta ordlistan över AI
Balanced Bootstrap
A variant of bootstrap ensuring that each original observation appears the same number of times across all bootstrap samples, reducing bias in error estimation.
Stratified Bootstrap
A bootstrap sampling method that preserves the proportion of classes or strata in each resampled sample, essential for imbalanced data.
Weighted Bootstrap
A technique where each observation is assigned a probability weight during bootstrap sampling, allowing oversampling of data points considered more important or reliable.
Double Bootstrap
An iterative procedure where bootstrap sampling is applied to another bootstrap sample, used to improve the accuracy of confidence intervals and correct higher-order biases.
Block Bootstrap
A bootstrap variant designed for dependent data (time series, spatial data) that resamples contiguous blocks of observations to preserve the correlation structure.
Bayesian Bootstrap
Interpretation of bootstrap as an approximation of a Bayesian posterior distribution, where bootstrap sampling simulates a non-informative prior distribution on the data.
Parametric Bootstrap
A method where bootstrap samples are generated from a parametric distribution fitted to the original data, unlike non-parametric bootstrap which uses the data themselves.
Non-Parametric Bootstrap
Standard bootstrap approach that makes no assumptions about the underlying data distribution, relying solely on direct resampling of the empirical sample.
Bootstrap Percentile Confidence Interval
Method for constructing confidence intervals using the percentiles (e.g., 2.5% and 97.5%) of the distribution of estimators calculated from bootstrap samples.
Bootstrap BCa Confidence Interval
Bootstrap confidence interval adjusted for bias (Bias-Corrected) and acceleration (accelerated), offering better accuracy by correcting biases and non-normality of the bootstrap distribution.
Bootstrap Subsampling (Subagging)
Variant of bagging using subsets of observations drawn without replacement, reducing variance while being less computationally expensive than traditional bagging.
Bootstrap for Time Series
Set of techniques (such as block bootstrap or residual-based bootstrap) adapted for resampling time series data while respecting their temporal dependence.
Jackknife-after-Bootstrap
Diagnostic method that uses the jackknife technique on bootstrap results to assess the stability and influence of each observation on the bootstrap estimation.