Přejít k hlavnímu obsahu

Přihlášení pro studenty

Přihlášení pro zaměstnance

Publikace detail

Bootstrap Method and Confidence Intervals
Autoři: Kubanová Jana | Linda Bohdan
Rok: 2005
Druh publikace: ostatní - přednáška nebo poster
Název zdroje: ISBIS4 Conference Abstracts
Název nakladatele: The international society for Business and Industrial Statistics
Místo vydání:
Strana od-do:
Tituly:
Jazyk Název Abstrakt Klíčová slova
cze Bootstrapová metoda a intervaly spolehlivosti V článku je řešen problém konstrukce intervalu spolehlivosti v případě, kdy nejsou splněny základní předpoklady, jako normalita rozložení. Ukazuje se, že je možné dosáhnout dobrých výsledků pomocí resamplingu. interval spolehlivosti, resamplingová metoda
eng Bootstrap Method and Confidence Intervals Very often practical problem of statistical analysis is to express the most reliable conclusion when the random sample is too small and distribution of used statistics is unknown and hardly conjecturable. It is very often in research in economics, sciences and technical sciences. The application of common, classical methods without fulfillment of required assumptions can cause that untrustworthy results are obtained and then it is very difficult to express any conclusions about standard errors, hypotheses testing or confidence intervals. The methods of modern statistical analyses can in some cases help to solve this problem. The methods are called bootstrap methods. Bootstrap methods are based on resampling new samples from the original data set with the same rate. This approach involves repeating the original data analysis procedure with many replicated datasets. Very important advantage of this method is that it allows constructing artificial data sets without making any assumptions about bell shaped curves. Problems that can be solved with bootstrap method can be divided into two groups that are later called parametric and nonparametric bootstrap. Parametric simulation assumes that distribution of random variable X is known. Nonparametric simulation does not require this assumption. Concrete examples demonstrate both ways of simulation. Bootstrap approach to confidence intervals construction Four methods of confidence interval construction based on simulated resampled samples are described in the paperThe important assumption of good results of above mentioned quantile method is an unbiased estimate of the parameter T. The common problem is that the form of distribution changes when F is replaced by /F/. Examples of bootstrap simulated confidence intervals In order to verify the accuracy and correctness of all above-mentioned ways of bootstrap confidence interval construction, 10 000 bootstrap replications from known distribution were made and confidence inter confidence interval, resampling method