|
Implementation and Application of Statistical Methods in Research, Manufacturing Technology and Quality Control
Kupka, Karel ; Šeda, Miloš (referee) ; Militký, Jiří (referee) ; Dohnal, Gejza (referee) ; Karpíšek, Zdeněk (advisor)
This thesis deals with modern statistical approaches and their application aimed at robust methods and neural network modelling. Selected methods are analyzed and applied on frequent practical problems in czech industry and technology. Topics and methods are to be benificial in real applications compared to currently used classical methods. Applicability and effectivity of the algorithms is verified and demonstrated on real studies and problems in czech industrial and research bodies. The great and unexploited potential of modern theoretical and computational capacity and the potential of new approaces to statistical modelling and methods. A significant result of this thesis is also an environment for software application development for data analysis with own programming language DARWin (Data Analysis Robot for Windows) for implemenation of effective numerical algorithms for extaction information from data. The thesis should be an incentive for boarder use of robust and computationally intensive methods as neural networks for modelling processes, quality control and generally better understanding of nature.
|
|
Bayesian statistical modelling
Vilikus, Ondřej ; Hebák, Petr (advisor) ; Berka, Petr (referee) ; Militký, Jiří (referee)
Conjoint analysis is a popular method in consumer preferences research. One of the factors that caused the increasing popularity of this method in recent years is wide use of hierarchical Bayesian models which has been found invaluable in solving the problem of how to obtain reliable estimates of individual preferences without need for overloading respondents with too many conjoint tasks. First goal of my dissertation was to confirm whether the use of Bayesian models is the best choice under all circumstances or whether there are some limitations of this approach. For this purpose I conducted a study based on simulated datasets. Algorithm used enabled generation of datasets that differed in several parameters of interest but which were most comparable in other aspects. Results show that hierarchical models represent choice leading to highest accuracy in predicting respondents' choices in holdout tasks. Use of hierarchical models is most beneficial in the situation of strongly heterogeneous population yet limited amount of available data. In these cases we are able to capture the structure of heterogeneity with significantly lower number of choice task necessary from each respondent. Second goal of the dissertation was to answer the question whether we can increase also the effectiveness of the questioning in conjoint analysis by adding several direct questions. Suggested hybrid choice-based conjoint method (HCBC) combines conjoint analysis tasks with direct questions regarding the preference of levels for each attribute. These are used during the estimation of the model and for increasing the effectiveness if the conjoint analysis tasks design. The HCBC was compared with traditional choice-based conjoint (CBC) and adaptive choice-based conjoint (ACBC) based on practical study involving 421 respondents randomly assigned in one of three test groups. Suggested method has been found as useful alternative that can help with reducing number of choice task needed and as a solution for some situations when diverse importance of the attributes tested does not allow for indirect estimation of preferences with respect to all attributes tested.
|
| |
| |
| |
| |
| |
| |
| |
| |