A Bootstrap Confidence Interval for Skewness and Kurtosis and Properties of t-test in Small Samples from Normal Distribution

Loading...
Thumbnail Image

Date

2009

Journal Title

Journal ISSN

Volume Title

Publisher

Aves Yayincilik, Ibrahim Kara

Access Rights

info:eu-repo/semantics/openAccess

Abstract

Objectives: We examined the test properties about mean and mean differences, sampling distributions properties of mean and standard deviation, especially when the sample size is lower than 10 and the variable has normal distribution in population. In addition, we aimed to construct a 95% bootstrap confidence interval for skewness and kurtosis values in various samples sizes. Materials and Methods: In our simulation study, 10,000 samples with replacement were taken from the standard normal population Various sample sizes were evaluated Data which were obtained from an animal study were used for comparison of t-test's and Wilcoxon sign test's power in small sample. Results: According to our results, sampling distribution of skewness coefficients has normal distribution, however, kurtosis coefficients have positively skewed distribution. A bootstrap confidence interval for these coefficients by using these sampling distribution can be used for normality test Moreover, it was shown that test statistic for the mean has t-distribution in all studied sample sizes Conclusion: We can say that distribution shape of the variable plays a more important role than sample size in selection of appropriate test statistic

Description

11th National Congress of Biostatistics -- MAY 27-30, 2008 -- Malatya, TURKEY
Ankarali, Handan Camdeviren/0000-0002-3613-0523; Ankarali, Seyit/0000-0003-3752-0846
WOS: 000272760100003

Keywords

Small sample size, bootstrap simulation, power of t-test, confidence interval, skewness and kurtosis

Journal or Series

Trakya Universitesi Tip Fakultesi Dergisi

WoS Q Value

Q4

Scopus Q Value

Volume

26

Issue

4

Citation