BEGIN:VCALENDAR VERSION:2.0 X-WR-CALNAME:Department of Statistics Seminar PRODID:-//Department of Statistics//NONSGML Stephen Cope//EN BEGIN:VEVENT DTSTART:20120216T160000 DURATION:PT60M SUMMARY:On Akaike and likelihood cross-validation criteria for model sele ction UID:2012-02-16-1600@vcal.stat.auckland.ac.nz DESCRIPTION:The talk discusses Akaike and likelihood crossvalidation crit eria for model/estimator choice. After a presentation of the main concep t on model selection\, we will focus on the choice of estimators in non- standard cases. First\, we study two examples arising when we wish to as sess the quality of estimators on a particular set of information\, whil e the estimators may use a larger set of information. The first example occurs when we construct a model for an event which happens if a continu ous variable is above a certain threshold. We can compare estimators bas ed on the observation of only the event or on the whole continuous varia ble. The other example is that of predicting survival based on survival information only\, or using in addition information on patient's disease . We develop modified AIC and LCV criteria to compare estimators in this non-standard situation. Second\, we study the choice of estimators in p rognostic studies. Estimators for a clinical event may use repeated meas urements of markers in addition to fixed covariates. These measurements can be linked to the clinical event by joint modelling involving latent structures. When the objective is to choose between different estimators based on joint models for prediction\, the conventional Akaike informat ion criterion (AIC) is not well adapted and decision should be based on predictive accuracy. We define an adapted risk function called expected prognostic cross entropy (EPCE) and further modify it for right-censored observations. The risk functions can be estimated by leave-one-out cros s validation\, for which we give approximate formulas and asymptotic dis tributions. LOCATION:ECE Seminar Room 303.257\, Science Centre CREATED:20120112T023603Z DTSTAMP:20120112T203259Z LAST-MODIFIED:20120112T203259Z SEQUENCE:2 END:VEVENT BEGIN:VEVENT DTSTART:20120220T110000 DURATION:PT60M SUMMARY:How big are the real mortality reductions produced by cancer scre ening? Why do so many trials report only 20%? UID:2012-02-20-1100@vcal.stat.auckland.ac.nz DESCRIPTION:Influential reports on the reductions produced by screening f or cancers of the prostate\, colon and lung have appeared recently. The reported reductions in these randomized trials have been modest\, and sm aller than expected. But even more surprisingly\, all three figures are very similar. I explain why these figures are underestimates and why the seemingly-universal 20% reduction is an artifact of the prevailing data -analysis methods and stopping rules. A different approach to the analys is of data from cancer screening trials is called for. LOCATION:ECE Seminar Room 303.257\, Science Centre CREATED:20120112T023227Z DTSTAMP:20120112T215527Z LAST-MODIFIED:20120112T215527Z SEQUENCE:3 END:VEVENT BEGIN:VEVENT DTSTART:20120222T110000 DURATION:PT60M SUMMARY:Delights of directional statistics: (a) free-lunch learning\, (b) crystals\, earthquakes and orthogonal axial frames UID:2012-02-22-1100@vcal.stat.auckland.ac.nz DESCRIPTION:Observations that are directions\, axes\, or rotations requir e the techniques of directional statistics. This talk aims to illustrate the special flavour of this area through glimpses at two topics.\n\n(a) Free-lunch learning\n Free-lunch learning (FLL) is a phenom enon in which relearning partially-forgotten mental associations induces recovery of other associations. When memory is modelled in terms of an artificial neural network\, the extent of FLL can be quantified in geome trical terms and involves Grassmann manifolds of subspaces of the weight space. Joint work with Jim Stone (Psychology\, Sheffield) will be descr ibed\, in which simple properties of uniform distributions yield results on the expected amount of FLL. The form of forgetting plays an importan t role.\n\n(b) Crystals\, earthquakes and orthogonal axial frames\n Orthogonal axial frames are (ordered) sets of orthogonal axes. They arise as (i) key geometrical elements (known in seismology as `focal mec hanisms') of earthquakes\, (ii) principal axes of certain physical tenso rs (e.g. stress tensors)\, (iii) axes of orthorhombic crystals. Some too ls for the analysis of data that are orthogonal axial frames will be wil l be described. This is joint work with Richard Arnold (Wellington).\n LOCATION:ECE briefing room 257 CREATED:20120114T053414Z DTSTAMP:20120202T020633Z LAST-MODIFIED:20120202T020633Z SEQUENCE:5 END:VEVENT BEGIN:VEVENT DTSTART:20120308T160000 DURATION:PT60M SUMMARY:Optimal Asset Pricing UID:2012-03-08-1600@vcal.stat.auckland.ac.nz DESCRIPTION:It is a well-known phenomenon that airline passengers travell ing on the same flight (same origin and same destination) and in the sam e class (cabin) will often have paid substantially different fares. This apparent anomaly in the pricing pattern is due to the fact there is a t ime-varying elasticity of demand (or "price sensitivity") for this parti cular "product".\n\nMy co-author Pradeep Banerjee and I have developed a differential equations model which permits one to derive an optimal pri cing policy in such a setting. (The policy is "optimal" in terms of the expected value of a stock of goods at a specified time.) Deriving the op timal policy requires a model for the price sensitivity and for an inhom ogeneous Poisson arrival rate of customers. So far we have worked with s mooth price sensitivity functions. However it is somewhat easier to tran slate intuitive conjectures about price sensitivity into a function fram ed as being piecewise linear in price.\n\nIn this talk I will explain a bit about how the differential equations for the optimal prices are deri ved\, and then discuss how the technique must be adjusted to deal with t he piecewise linear setting. I will also discuss some of the techniques that I and my Summer Scholarship student Ray Shahlori have used to code up the solution procedure in R. I will show some examples of solutions . LOCATION:Statistics Seminar Room 303.279\, Science Centre CREATED:20120112T204154Z DTSTAMP:20120131T222645Z LAST-MODIFIED:20120131T222645Z SEQUENCE:6 END:VEVENT END:VCALENDAR