Jostein Lillestøl

Bayesian Estimation of NIG-parameters by Markov Chain Monte Carlo Methods

The Normal Inverse Gaussian (NIG) distribution recently introduced by Barndorff-Nielsen (1997) is a promising alternative for modelling financial data exhibiting skewness and fat tails. In this paper we explore the Bayesian estimation of NIG-parameters by Markov Chain Monte Carlo Methods.

MORTALITY PATTERNS FOLLOWING INTERNAL FIXATION FOR ACUTE FEMORAL NECK FRACTURES IN THE ELDERLY WITH SPECIAL EMPHASIS ON POTENTIAL EXCESS MORTALITY FOLLOWING REOPERATIONS

Mortality patterns and excess mortality have been studied and quantified in 103 patients treated with internal fixation for acute, displaced femoral neck fractures with special emphasis on the potential excess mortality which may follow later operations for capital necrosis, failure of the osteosynthesis, etc. Of 103 patients studied 31 needed one or more reoperations. We have confirmed previously published reports that excess mortality is limited to the first six months after the primary operation. Quantification of the excess mortality which may follow reoperations shows that later operations are not followed by an increased death rate compared with the standard population.

MIGRATION OF THE FEMORAL STEM IN HIP ARTHROPLASTIES Analysis of Associations with Structural, Radiological and Follow-up Variables

Medical and/or distal migration of the prosthetic stem was found in 63 out of 337 patients (19 per cent) treated surgically with a Christiansen hemiprosthesis, a Christiansen total hip prosthesis or a Charnley total hip prosthesis. Fractured bone cement, radiolucent zones at the cement/bone interface, resorption of the femoral calcar and cortical sclerosis were all associated with migration. Varus position of the Christiansen total hip prosthesis was significantly associated with medical migration, and a short stem was significantly associated with distal migration. The other structural variables could not be linked with migration. Distal migration was pain-inducing and was significantly associated with late infection. Medical migration had a less distinct association with pain, and was not correlated with infection. Both medical and distal migration were time-dependent, and 4 or more years after operation about 2.5 per cent of the prosthetic stems had migrated.

Multivariate Modelling and Prediction of Hourly One-Day Ahead Prices at Nordpool

In this paper, we exploit multivariate and functional data techniques to capture important features concerning the time dynamics of hourly one-day ahead electricity prices at Nordpool. The methods are also used to obtain pragmatic prediction schemes for these prices. Following Huisman et al. (Energy Economics 29, 2007), we first consider the 24-hourly one-day ahead prices as a 24-dimensional variable observed on a daily basis. These are analyzed by means of a multivariate analysis of variance (MANOVA) and the results are presented by some enlightening graphs. We then account for the smoothness of the curve that these 24 values form when they are seen as a function of time of the day. This smoothness is exploited by a functional analysis of variance (FANOVA). Multivariate prediction is then addressed, first by a simple univariate ARIMA scheme combined with hourly corrections obtained from the MANOVA and then by a true multivariate state space scheme.

Some crude approximation, calibration and estimation procedures for NIG-variates

In this paper we explore some crude approximation, calibration and estimation procedures for Normal Inverse Gaussian (NIG) variates of potential use in risk management. Among others we treat in some detail the calibration of bivariate NIG consistent with marginal NIG.

Survival patterns following primary prosthetic replacement for acute femoral neck fractures in the elderly.

Theoretical statistical models have been used to study survival patterns following primary prosthetic replacement for acute femoral neck fractures in the elderly with special emphasis on the nature of the potential excess mortality following trauma and operation. The chances of our patients surviving given periods of time are well below those of the standard population. The excess mortality has been quantified using the proportional hazard model. The high death rate for our patients is limited to the first 4–6 postoperative months; thereafter the death rate is similar to that of the standard population. Similarly we can show that mortality following surgery is primarily dependent on the age of the patient. Survival probability and expected longevity have been calculated for different age groups. The relative longevity expected for patients below the age of 70 is very good compared with a very reduced expected longevity for patients older than 90 years. The practical consequences of our findings are discussed.

Multivariate statistical methods for quality creation: a review

This paper discusses the role of multivariate statistical methods in quality creation, i. e. for exploring customer needs, product quality deployment and management planning. A brief survey of methods are given from different ordering perspectives. Some thoughts are given to (i) common methods in marketing and consumer research (ii) matrix data analysis as one of ‘the seven management and planning tools’. At the end an example from Kano et al. concerning trends of quality elements is reanalysed by correspondence analysis and a perceptual map with some novel features is given.

Statistical testing of bounded rationality with applications to the newsvendor model

In this paper we define bounded rationality in terms of probabilistic cost efficiency. This establishes a framework where bounded rationality can be examined rigorously by statistical methods. We apply this theory to data from laboratory experiments on the newsvendor model. We enter into a theoretical discussion of the classical pull-to-center effect and explain why underorders are typically larger than overorders.

Creaming and the Likelihood of Discovering Additional Giant Petroleum Fields

The applied context of this paper is the exploration for petroleum resources, like petroleum accumulations in different deposits, where the most promising deposits are likely to be drilled first, based on some size indicators, so-called creaming. The paper explores creaming models in the context of sampling with probabilities in proportion to size, for which a lognormal size distribution has nice analytical features. It departs from the traditional paradigm in petroleum resource assessment. Instead of conceiving a finite population being depleted over time in a decaying fashion with respect to size, the situation is studied within the framework of independent observations (infinite population) and an exploration maturity-dependent creaming factor. The theoretical and practical consequences for inference on the parent population and the probabilities and expectations linked to future discoveries are studied. The theory applies to the issue of remaining sizes of petroleum resources to be found within different future discovery horizons on the mature part of the Norwegian Continental Shelf. The aim is to obtain reasonable and useful predictions, and not to provide the best possible explanation of the exploratory behavior itself.

The Z-Poisson distribution with application to the modelling of soccer score probabilities

This paper presents a non-standard multivariate Poisson-type distribution, here named as Z-Poisson distribution. Some details on the bivariate case are worked out in the context of modeling soccer (football) scores, and the distribution is fitted to score data from the top league in the 2007–08 season for 10 European countries.