Conference Programme 2015
Tuesday 14th July Wednesday 15th July Thursday 16th July Friday 17th July
Tuesday 14th July, 11:00 - 12:30 Room: HT-101
Advanced statistical modeling
|Convenor||Dr Jarl Kampen (Research Methodology Group, Wageningen UR, The Netherlands )|
Paper Details1. A Poisson Regression Model Handling Truncated Data with Multiple Inflations
Dr Ting Hsiang Lin (NTPU)
Dr Min-hsiao Tsai (NTPU)
In this study, we extended the zero-inflated Poisson regression, and used number of days of a behavior in two weeks as example. The events can be classified into inflated values and non-inflated values; the former can be considered as multinomial distribution, and the later is a truncated Poisson distribution. We proposed a multiple-inflated truncated Poisson regression model, and the model is a mixture of the multinomial logistic regression and the truncated Poisson regression.
2. A simple test to substantiate the claim that ordinal variables are the product of underlying bivariate normal distributions
Dr Jarl Kampen (Research Methodology Group, Wageningen UR, The Netherlands)
Dr Arie Weeren (Statua, Antwerp University, Belgium)
A simulation study was carried out to study the behavior of the polychoric correlation coefficient in datasets which data generating process was incompatible with the underlying variable paradigm. We provide evidence that the hypothesis that observed ordinal variables are crude measures of normally distributed underlying variables can be verified by a simple test, which under certain conditions may serve as evidence of underlying variables. If underlying bivariate normality is rejected, what alternative DGP is responsible for observed data remains undecided. Impotence to find falsifying evidence for underlying variables justifies doubt about the scientific value of theories based on their postulation.
3. Continuous functions: Approximation via Sampling and Coefficient of Variation
Professor Nikolaos Farmakis (Aristotle University of Thessaloniki, School of Mathematics)
Ms Ioanna Papatsouma (Aristotle University of Thessaloniki, School of Mathematics)
A basic, very useful parameter for the study of the random variables(rv) is the Coefficient of Variation(Cv). It is used already for many Statistical applications, e.g. bias’ correction, approximation of distributions’ polynomial form, etc.
A new use is proposed: Having a continuous curve with about symmetric shape defined in I=[0, b] or generally in I=[a, b], we collect a sample of values ym=f(xm), xm ∈I, in order to get a polynomial approximation of the curve’s function f(x),x ∈I. We arrive to this resulting function via an algorithm based on Cv,