Pasikliautinųjų intervalų skaičiavimo būdų palyginimas diskretiems biologiniams duomenims

Abstract

The bootstrap (resampling) methods of construction of confidence intervals (c.i.) are considered. The c.i. obtained are comapred with each other and with the standard student c.i. by simulation. The study is based on a real biological data about distribution of badger, fox, and raccoon dog borrow number in 1 km² square areas uniformly allocated throughout Lithuania territory. The distributions take only a few nonzero values with relatively large proportion of zeros, are very skewed and have irregular tails. In this case the normal approximation, the student c.i. is based on, of the distribution of the sample mean is not sufficiently accurate for small samples.