|
|
Statistical Analysis of Wiener Process Based on Partial Observations
Hrochová, Magdalena ; Hlubinka, Daniel (advisor) ; Omelka, Marek (referee)
Wiener process-a random process with continuous time-plays an important role in mathematics, physics or economy. It is often good to know whether it contains any deterministic part, e.g. drift or scale. However, it is nearly impossible either observe the whole trajectory of the process or preserve its full history. This thesis deals with a statistical analysis based on partial observations, namely passage times through some given barriers. We propose several statistical methods for testing hypotheses about drift or scale using these observations. As supporting methods, we consider the maximum likelihood theory, non-parametric test against a trend, and binomial test. For testing the value of scale in the model with no drift and constant scale we recommend maximum likelihood theory. We derive the estimate and related tests in the case of observing only three barriers. The simulation study suggested observing more barriers for testing monotony of scale in a model with linear drift, or testing monotone and convex/concave drift in a model with constant scale. 1
|
|
|
Levene's Test for Equality of Variances
Hrochová, Magdalena ; Omelka, Marek (advisor) ; Kulich, Michal (referee)
In the presented bachelor thesis, we focused on the Levene's test and its modifications that are used to assess the equality of variances for k independent random samples. At the beginning, we described Analysis of variance (ANOVA) that is a method for analyzing differences between group means for k independent random samples. The following part contains the original Levene's test description, including a discussion about the assumptions' verification for using ANOVA. Considering the fact that the original test is not confident in case of samples from specific distributions (e.g. a chi-square distribution) we summarize some known and suggest some new modifications. The main part of the thesis is dedicated to experimental simulations. In the conclusion, we discuss the simulation results for different versions of the Levene's test and other similar tests for equality of variances. 1
|
guest ::
RSS feed.