Science, Technology and Engineering

Title: Lump, don't split: improved significance tests for model validation and environmental monitoring

Speaker: Dr. Andrew Robinson, University of Melbourne

Time & date: 2 p.m., Friday, 3rd November 2006

Venue: Room 310, Physical Sciences 2 Building (Access Grid Room), La Trobe University, Bundoora Campus

Abstract: The sustainable maintenance of Australia's biodiversity requires statistically valid monitoring and modeling programs. This need is poorly met by current statistical tools, because the current tools focus on finding differences (splitting) rather than identifying similarities (lumping). Equivalence-style hypothesis tests can fill this important need, because they lump instead of splitting. I will introduce, review, and compare the available statistical tooks for equivalence tests.

 

---

 

Title: Calibrated Empirical Likelihood Estimation using a Displacement Function: Sir R.A. Fisher’s Honest Balance

Speaker: Sarjinder Singh, Department of Mathematics and Statistics, La Trobe University

Time & date: 2 p.m. Friday 6th October 2006

Venue: Room 416, Thomas Cherry Building, La Trobe University, Bundoora Campus

Abstract: In the present investigation, we propose a new method to calibrate the estimator of the general parameter of interest in survey sampling. We demonstrate that the linear regression estimator due to Hansen, Hurwitz and Madow (1953) is a special case of this. We reconfirm that the sum of calibrated weights has to be set equal to the sum of the design weights within a given sample as shown in Singh (2003, 2004, 2006), Stearns and Singh (2005) and Singh and Arnab (2006). Thus, this demonstrates that Sir. R.A. Fisher’s brilliant idea of keeping the sum of observed frequencies equal to that of expected frequencies leads to a “Honest-Balance” while weighing design weights in survey sampling. The major benefit of the proposed new estimator is that it never fails like the pseudo empirical likelihood estimators listed in Owen (2001). The main endeavor of this paper is to bring a change to the existing calibration technology, which is based on only positive distance functions, with a displacement function that has the flexibility of taking positive, negative, or zero value. At the end, the proposed technology has been compared with its competitors under several kinds of linear and non-linear non-parametric models using an extensive simulation study. This paper will encourage scientists/researchers to think more on these lines.

 

---

 

Title: Biomarkers and Robustness

Speaker: Prof. Stephan Morgenthaler, Ecole Polytechnique Federale, Lausanne, Switzerland

Time & date: 2 p.m., Friday 11th August 2006

Venue: Room 310, Physical Sciences 2 Building (Access Grid Room), La Trobe University, Bundoora Campus

Abstract: The talk explains biomarkers and their uses, in particular when trying to find genetic causes for complex diseases. A method is described to discover if one or more genes carries one or more alleles each conferring risk for any common disease. The method does not depend upon genetic linkage of risk-conferring alleles to high frequency genetic markers such as single nucleotide polymorphisms. Instead, the sums of allelic frequencies in cohorts comprised of persons with and without the disease are determined and a statistical test is applied to discover if the difference in sums observed is greater than would be expected by chance.

A statistical model is presented that defines the ability of such tests to detect significant gene/disease relationships as a function of sample sizes and values of key confounding variables such as multigenic risk, errors in diagnosis and allele recognition, ethnic diversity in the general population and the expectation that most alleles detected and enumerated will be neutral with regard to disease risk.

If time permits, the robustness aspects of such tests are also considered, in particular the effects of deviations from the standard model on the power.

 

---

 

Title: An introductory guide to statistical programming using Matlab

Speaker: Mr Scott Manderson, Department of Mathematics and Statistics, La Trobe University.

Time: 2.00pm, Friday 16 June, 2006.

Venue: Access Grid Room, Level 3, Physical Sciences 2, La Trobe University, Bundoora Campus.

Abstract: Due to the massive increase in computer processing power over the last decade or two, a greater emphasis appears to have been placed on comparing finite sample Monte Carlo simulations, rather than asymptotic results. With this increase in processing power, a wide range of competing computer software has become available on the market, designed specifically for this purpose. It is therefore difficult for prospective researchers to initially make the right choice of which software will best suit their needs, and changing from one language to another mid-way through ones research is never ideal.

For this reason I present a beginners guide to programming using Matlab. This seminar is aimed particularly at students and researchers who are interested in using Matlab as an analytical tool for statistical research. The basics of using Matlab and some of its potential will be discussed. The seminar assumes no prior knowledge of Matlab, although prior knowledge with programming in general will be an advantage. Previously S-Plus and R have been the common choices for researchers in this area, so I also highlight similarities between using Matlab and these two programs in particular.

 

---

 

Title: A non-parametric test for self-similarity and stationarity in network traffic.

Speaker: Dr. Owen Jones, Department of Mathematics and Statistics, University of Melbourne.

Time: 2 - 3 pm, Friday, 31 March, 2006.

Venue: Room 416, Thomas Cherry Building, La Trobe University, Bundoora Campus.

Abstract: During the last decade packet traces collected from both Local Area Networks (LAN) and Wide Area Networks (WAN) have been extensively analysed in the framework of self-similar processes. Not only is there ample empirical evidence supporting the existence of self-similarity in network traffic data, but also mathematical models have been established which suggest physical explanations for observed self-similarity. Of course in practice self-similarity can only hold over a finite range of scales, whence the need for a statistically founded test for determing the scaling range. A further difficulty that arises when modelling packet traces is distinguishing long-range correlation from trends in the mean packet arrival rate.

In this talk we present a non-parametric statistical test for self-similarity and use it both to determine scales over which we have a constant scaling regime and to determine time intervals wherein the intensity process appears to be stationary. The method is then applied to some publicly available LAN and WAN traces. These traces all exhibited self-similarity over a range of around 5 seconds to 5 minutes. The change in scaling below 5 seconds is most likely due to the effect of the network protocols used to transmit packets. Above 5 minutes we may be seeing the effect of finite network capacity, which puts an upper limit on the traffic intensity. Some of the traces we looked at also showed changes in the mean packet arrival rate and/or scaling behaviour over time. The length of stationary periods varied from 5 minutes to 120 minutes (the longest trace we considered).